You are on page 1of 117

MODAL ANALYSIS OF MEMS GYROSCOPIC SENSORS

by

Marc Burnie

A thesis submitted to the Department of Mechanical & Materials Engineering

In conformity with the requirements for

the degree of Master of Applied Science

Queen’s University

Kingston, Ontario, Canada

(May, 2010)

Copyright ©Marc Burnie, 2010

i

Abstract

Microgyroscopes find popular applications in modern life, such as, vehicle navigation, inertial

positioning, human body motion monitoring, etc. In this study, three unique MEMS gyroscopic

sensors were investigated using experimental methods and finite element analysis (FEA)

modelling, particularly their modal behaviour. The analytical, simulated and experimental results

were compared and the discrepancy between resonant frequencies of the significant mode shapes

was discussed.

Three microfabricated gyroscopes were investigated: a thermally-actuated in-plane gyroscope, an

electrostatically-actuated in-plane gyroscope and an electrostatically-actuated out-of-plane

gyroscope. Numerical finite element modal analysis for these three gyroscopes was conducted

using COMSOL MultiphysicsTM. The experimental testing was conducted using a microsystem

analyzer (MSA-400 PolyTec) with an integrated laser vibrometer.

The simulation models predicted that the frequencies for driving and sensing modes were

4.948kHz and 5.459kHz for a thermally-actuated gyroscope, which agreed well with

experimentally determined results of 5.98kHz and 6.0kHz respectively. The power requirements

of a thermally-actuated gyroscope were 363.39mW to elicit a maximum peak-to-peak

displacement of 4.2µm during dynamic operation. Similarly, the simulated frequencies for the

driving and sensing modes were 1.170kHz and 1.644kHz for an electrostatically-actuated in-plane

gyroscope, which corresponded to experimentally determined resonant frequencies 1.6kHz and

1.9kHz.

Simulation for the electrostatically-actuated out-of-plane gyroscope was conducted and the

frequencies for the driving and sensing modes were found to be 2.159kHz and 3.298kHz. Due to

ii

some fabrication defects, the experimental testing for this microgyroscope was not successful.

Some recommendations to improve the design were provided for the future work.

iii

iv . Dedication I would like to dedicate this thesis to my father. who taught me the value of education and the meaning of hard work. Martin Burnie.

I would also like to thank Jacky Chow. v . I would like to thank my parents and the rest of my family for their ceaseless support. Acknowledgements I would to sincerely thank my supervisor. Xin Guo and Kongying Xie. for his guidance. my lab mates. Dr. I would not have been able to complete this thesis if it wasn’t for his valuable advice and support. Yongjun Lai. understanding and patience. for their assistance with lab equipment and being great people to work with in general. Finally.

........................................................................... 9 1............................................................... Table of Contents Abstract ......................................................................................................................................................... ii Dedication ................................................................................................................................................................................ iv Acknowledgements ............ 1 1.......................1 Applications ..................................................................................................................3.............................. 3 1................2 Deposition .....................................................................................................................................................................1 Introduction to Micro-Electro-Mechanical Systems .................................................................................. 17 2............ 15 Chapter 2 Literature Review ...1 Electrostatic Actuation ..................................................................................................3..................... 4 1..................................................................... 1 1...3 Gyroscope Kinematics .......................................................................................................................................................................................................4 Performance Characterization ................................................ 7 1............4 Fabrication Fundamentals .......................................................1....................................................................................1.......1 Etching ..........2................................2 Operating Principles ................... vi List of Figures ................................................................................ 14 1............................. 19 vi ............................................................................2............................................................................. 2 1......................................................................................2......................2 Thermal Actuation ................................ ix List of Tables ....................................................... v Table of Contents ......................................... 11 1.....................1 Lithography Modes ........................................................ 16 2.................... xiii Symbols and Nomenclature ............................................................ 16 2.................... 9 1.........................................................................................................................2 Micro-gyroscopes .......2................................................................................. 2 1.5 Thesis Outline ........................ xiv Chapter 1 Introduction ..........................................................................................3 MEMS Actuators .....................

...................................1..................2 Microfabrication Processes ............................................................. 44 3......... 77 4........2. 78 vii .......... 66 4........................ 51 3..........1.............. 69 4.......................1.......... 24 2..............................................................................................................................................1... 36 3.... 43 3............................................3 Thermally-Actuated Gyroscope Simulation Results ...........1 Thermally-Actuated Gyroscopic Sensor Analytical Model ............. 20 2......................4 Anodic Bonding ................................ 72 4..........1 Thermally-Actuated Gyroscope Experimental Testing ........................3 Thermally-Actuated Gyroscope Static Measurements ...........................................................................................................................2 Micragem Silicon-On-Insulator (SOI) ...............5 Electrostatically-Actuated SOI Gyroscope Simulation Results ........... 66 4.................................................................................1.....................5 Thermally-Actuated Gyroscope Planar Sensing Mode Resonant Frequency Characterization .........................2 Thermally-Actuated Gyroscope Prototype .......................................................................................................................3 Previous Micro-gyroscope Research.........1 Metal Multi-User MEMS Process (MUMPS)............................................................1 Analytical Model Assumptions and Constraints ................................................................................... 70 4.............................................. 66 4..... 38 3....................................................................................................1.1..............................2 Finite Element Analysis ....... 23 2........1..........3 Electroplating .................1..... 26 2....................... 58 Chapter 4 Experimental Determination of Modal Frequencies ............................... 36 3..............2 Electrostatically-Actuated Gyroscopes Experimental Testing ..............................................2.......................................................................... 24 2........................6 Thermally-Actuated Gyroscope Out-of-Plane Characterization ...1....1 Thermally-Actuated Gyroscope Testing Methodology................................................... 45 3..................................2 Spring Modelling ........................ 28 Chapter 3 MEMS Gyroscope Theoretical Modal Analysis ..... 2....................4 Thermally-Actuated Gyroscope Planar Driving Mode Resonant Frequency Characterization ...............3 Analytical Resonant Frequencies Calculation.... 37 3.1...4 Electrostatically-Actuated MetalMUMPS Gyroscope Simulation Results .. 75 4.................................................

.............................2 Electrostatically-Actuated MetalMUMPS Gyroscope Predicted Mode Matching ...........1 Thermally-Actuated Micro-gyroscope ................................... 78 4.... 102 viii ........................................................ 88 5.............................................3...................................2..............................2 Comparison of Simulated and Experimental Results......................................... 94 5.............2 Electrostatically-Actuated MetalMUMPS Micro-gyroscope .........................................................................3 Electrostatically-Actuated MetalMUMPS Gyroscope Planar Driving Mode Resonant Frequency Characterization ... 91 5........... 95 Chapter 6 Conclusion ..............................2....... 79 4........ 100 Appendix ....................................... 89 5...................................................................................2........................... 80 4....................2................................................................................ 94 5.......3 Performance Analysis ........ 94 5............3.........................3 Electrostically-Actuated SOI Gyroscope Prototype...... 87 5..................................3.......2...................................................... 88 5.........................................................................................1 Future Work Recommendations ....2 Thesis Contribution . 4........................................................................................1 Thermally-Actuated MetalMUMPS Gyroscope Mode Matching .....................3 Electrostatically-Actuated SOI Gyroscope Predicted Mode Matching ............................... 97 6.2.................. 98 6........................... 85 Chapter 5 Simulated and Experimental Results Analysis .....4 Thermally-Actuated Gyroscope Power and Efficiency ........3 Error Explanation ............1 Detrimental Mode Shapes .................2 Electrostatically-Actuated MetalMUMPS Micro-gyroscope Prototype ..............2....................................... 99 References ...4 Electrostatically-Actuated MetalMUMPS Gyroscope Out-of-Plane Characterization .................................................................................................. 87 5.........................................1 Electrostatically-Actuated Gyroscopes Testing Methodology ........................................................................ 95 5.................................. 83 4............................3..........

............................ .........Photolithography process steps.............3D view of an electrostatic gyroscope using the Metal MUMPS fabrication process........ 30 Figure 2-12 .... ..... 11 Figure 2-1 ...... ........................................................ 18 Figure 2-3 .........................Kinematics of a gyroscopic sensor system ...Drive mode beam deflection (left) and sensing mode deflection (right).................... .......SEM image of a comb-drive finger fabricated with the electroplating process........ b) A micrograph of the proposed gyroscope.........................................................................Hypothetical vibratory gyroscope...........SEM image of a robust 3 DOF mechanically decoupled gyroscope...................................................3D view of an electrostatic gyroscope using the Micragem process.................................................................................... ...........MEMS gyroscope utilizing independent drive and sense beams for mode decoupling....................................... ......................... List of Figures Figure 1-1-1 ................................................................................................... ... 26 Figure 2-8 .... 34 ix .....A planar...................Cross-sectional diagram of an electroplated nickel comb-drive resonator............... 21 Figure 2-4 .......................................... coupled micromachined vibratory gyroscope....... .............. 22 Figure 2-5 – Diagram of anodic bonding between a borosilicate glass and silicon wafer...........Cross-sectional view of an electrostatic gyroscope fabricated with the Metal MUMPS process................. 23 Figure 2-6 .... 25 Figure 2-7 .... .......... 4 Figure 1-1-2 .. 33 Figure 2-15 .......................Parallel plate actuator.........a) Conceptial diagram of a dual input axis gyroscope....................................................... ..................... 32 Figure 2-14 ......................... 28 Figure 2-10 ........................... ...........A novel double-gimbal micro-gyroscope by Draper Labs ...............................................Cross-sectional views of an electrostatic gyroscope using the Micragem process.......... ....Anisotropic (left) and isotropic (right) wet etching.... .... 29 Figure 2-11 .. 5 Figure 1-1-3 .............Micrograph of an in-plane Comb Drive Actuator.... ........................... 27 Figure 2-9 ................................................................................ 31 Figure 2-13 ................... 17 Figure 2-2 ............... 10 Figure 1-1-4 .............................

.........Thermally-actuated gyroscope simulated 1st mode at 3....... 57 Figure 3-19 ...... ..................... .............................159kHz....................................459kHz.Thermally-actuated gyroscopic sensor in-plane simplified model..............Thermally-actuated gyroscope simulated 5th mode at 8...................Electrostatically-actuated MetalMUMPS gyroscope simulated 2nd mode at 1. .....941kHz................................... .......316kHz................................. 52 Figure 3-14 .................. ........................ 38 Figure 3-5 ....510kHz............ ........................... 55 Figure 3-17 .. .................Thermally-actuated gyroscope simulated 2nd mode at 4......................... 53 Figure 3-15 ........Thermally-actuated gyroscope simulated 3rd mode at 5........................923kHz..............Analytical modelling of the flexures.................................................................. 54 Figure 3-16 ..........................500kHz...Electrostatically-actuated MetalMUMPS gyroscope simulated 1st mode at 1....... ...........Decoupling frame forces in 𝜑 mode shape....Folded flexure simplified model...................................... .......................Figure 2-16 – Frequency response curves of a 3 DOF gyroscope..... ............................................................... 41 Figure 3-7 .................... .Out-of-plane thermally-actuated gyroscopic sensor model......................Electrostatically-actuated MetalMUMPS gyroscope simulated 7th mode at 4..................... ..................................644kHz.........................Electrostatically-actuated MetalMUMPS gyroscope simulated 11th mode at 8.......................... ....948kHz...Electrostatically-actuated SOI gyroscope simulated 1st mode at 2.................................... ..Electrostatically-actuated MetalMUMPS gyroscope simulated 4th mode at 2.................................................... 37 Figure 3-3 .. .............................. . 49 Figure 3-12 ..................................................................................... .. 50 Figure 3-13 ...............Thermally-actuated gyroscope simulated 4th mode at 6...................................................... 35 Figure 3-1 .. 39 Figure 3-6 ................................ 59 x .....................Electrostatically-actuated MetalMUMPS gyroscope simulated 3rd mode at 1.......Thermally-actuated gyroscope FEA boundary condition definition diagram...................... 38 Figure 3-4 ...........246kHz............Decoupling frame simplified model.. .......... 56 Figure 3-18 .............. 48 Figure 3-11 ............ 47 Figure 3-10 .......................... 46 Figure 3-9 ........................... 41 Figure 3-8 .................. 36 Figure 3-2 ............................669kHz....170kHz......................................................................... ..............................

.........................454kHz.............. 65 Figure 4-2 ....Electrostatically-actuated SOI gyroscope simulated 4th mode at 3.................. 73 Figure 4-9 ........ ...... 61 Figure 3-22 ....Figure 3-20 ..... ............................... ................Electrostatically-actuated SOI gyroscope simulated 5th mode at 5........ 63 Figure 3-24 .... .................... ...................500kHz.... 70 Figure 4-6 ............. D) 5th observed mode at 8....... ..................679kHz..........Polytec MSA-400 and testing setup.Electrostatically-actuated SOI gyroscope simulated 6th mode at 5....................................... 79 Figure 4-14 – SEM image of an electrostatically-actuated MetalMUMPS gyroscope prototype...................Diagram of the electrical relationships between testing equipment apparatuses used for the thermally-actuated gyroscope.6kHz........... 67 Figure 4-1 ..... .. 60 Figure 3-21 .Thermal actuator displacement and current relationship................... ...................................................... 68 Figure 4-4 ..............................9kHz..Electrostatically-actuated SOI gyroscope simulated 7th mode at 7.....Polytec PMA reference image example......... 71 Figure 4-7 ....................Frequency response using a 225mARMS sinusoidal excitation signal at three frequencies..Experimentally determined mode shapes for a thermally-actuated gyroscope at A) 1st observed mode at 2..476kHz. ...1kHz and 6....... ..................................... 78 Figure 4-13 .. B) 2nd observed mode at 5.... ...........Laser interferometry equipment diagram...................Displacement of the thermal actuator at 3kHz excitation for different currents.......................... C) 3rd and 4th observed mode at 6......... .Electrostatically-actuated SOI gyroscope simulated 3rd mode at 3.... ................................... 80 xi ...... 64 Figure 3-25 ...................................298kHz.....................SEM image of an electro-thermally actuated MEMS gyroscope prototype... 72 Figure 4-8 ..Power consumption of the thermal actuator at several displacements elicited by changes in operational current... 67 Figure 4-3 ..... ...................................... ...Experimental out-of-plane FFT of the thermally-actuated gyroscope.. ..Diagram of the electrical relationships between testing equipment apparatuses used for the electrostatically-actuated MetalMUMPS gyroscope........ 69 Figure 4-5 ... .....494kHz............5kHz............................ 74 Figure 4-10 – Frequency response of the thermal gyroscope proof mass in the sensing direction........................4kHz.......Electrostatically-actuated SOI gyroscope simulated 2nd mode at 3..... ................76 Figure 4-11 ......... 62 Figure 3-23 . 77 Figure 4-12 .......

.........Frequency response of the electrostatic gyroscope proof mass in the driving direction....... C) 11th mode at 10.......................... 83 Figure 4-18 ........ 82 Figure 4-17 ......... 86 xii .............Figure 4-15 . B) 7th mode at 7................................................3kHz..............7kHz......Frequency response of the electrostatic gyroscope proof mass in the sensing direction.............. 84 Figure 4-19 – SEM image of an electrostatically-actuated MetalMUMPS gyroscope prototype: SOI design .................. ....................................... 81 Figure 4-16 ................................................................................1kHz and D) an out-of-plane mode at 12............................ ....................................................................................Experimental out-of-plane FFT of the electrostatically-actuated gyroscope.........................................................................................9kHz ...................................Experimentally determined mode shapes for a thermally-actuated gyroscope at A) 3rd mode at 2....................... ...

...........Electrostatically-actuated MetalMUMPS gyroscope mode shape comparison..............Thermally driven microgyroscopic sensor analytical and simulated summary of results........ List of Tables Table 3-1 ..... ... 51 Table 5-1 .... 90 xiii ................................................................................................................... ....................Thermally-actuated gyroscope mode shape comparison...... 89 Table 5-2 ...............

Abformung MEMS Micro-Electro-Mechanical Systems MUMPS Multi-User MEMS Process PMA Polytec Motion Analyzer PSV Polytec Scanning Vibrometer PVD Physical Vapour Deposition RMS Root-Mean Square SCSi Single-Crystal Silicon SEM Scanning Electron Microscope SOI Silicon-on-Insulator ZRO Zero Rate Output xiv . Galvanoformung. Symbols and Nomenclature Abbreviation Phrase AC Alternating Current CMOS Complementary Metal-Oxide-Semiconductor COMSOL Multiphysics Modelling Software CVD Chemical Vapour Deposition DC Direct Current DOF Degrees-of-Freedom DRIE Deep Reactive Ion Etching FEA Finite Element Analysis FEM Finite Element Modelling FFT Fast Fourier Transform ICs Integrated Circuits LIGA Lithografie.

The expense of exotic materials such as gold are negligible compared to cost of the microfabrication process. even identical devices that can be used as redundancy. and electronics on a silicon substrate through microfabrication technology. the price of fabrication is highly susceptible to economies of scale since the initial capital cost is higher than the material cost. Economically. which studies micro-scale physics and microfabrication processes for the creation miniature devices. actuators. The price of the equipment is reduced through the reapplication of outdated CMOS equipment and lithography masks can be used multiple times. Chapter 1 Introduction 1.  Low power – Less work is done to MEMS devices because smaller displacements and momentum forces are required. However. Most microfabrication processes are parallel by nature so many devices can be made for approximately the same cost as making one.1 Introduction to Micro-Electro-Mechanical Systems Micro-Electro-Mechanical Systems (MEMS) is the integration of mechanical elements. The heat capacity of micro-components is also relatively 1 . it is then reasonable to include many devices on a single wafer.1 In the lecture he identified a few features of this technology:  Lower Cost – The relatively small volume of the devices reduces the cost of materials. sensors. Richard Feynman (1962) realized the potential benefits of creating devices in the micro-scale when he conceptualized the MEMS field in his famous lecture “There’s Plenty of Room at the Bottom”. MEMS devices typically range in size between sub-microns to sub-millimeter. The intrinsic material value of each device is relatively low so they can be disposed of with little environmental impact.

With the advent of MEMS micro-gyroscope technology the range of cost-efficient applications of gyroscopic sensors has expanded. such as gravity and momentum. Scaling laws verify that the influence of volumetric forces. as well as diction of yaw.2.1 Applications Gyroscopic sensors are already an integral component of the automotive and aerospace industries with specific applications in roll-over detection in vehicles. 1. The integration of MEMS devices with IC technology decreases the required length of the electrical interconnects that ultimately reduces the power due to heat loss in the wires. small in comparison to human scale objects. pitch and roll orientation in missiles and aircraft.This was derived from the underlying forces that dominate in the micro scale. while surface area effects. such as surface tension and Van der Waals attraction. which allows for the use of joule heating as a method of actuation. The driving forces behind this trend are smaller 2 .  Fast response – The reduction in mass reduces the time and energy required to change state. Thermal and electrical components have less capacitance and mechanical components have higher resonant frequencies and rapidly accelerate. have on the devices at that scale would be reduced. MEMS devices have the ability to detect and react to disturbances much faster making them ideal for sensing applications. orientation diction in global positioning systems (GPS).  Insignificance of inertial forces .2 Micro-gyroscopes 1. would dominate.

greater power efficiency and lower cost.form-factors. the non- accelerating of reference. It is unreasonable to 3 . Some of the growing applications of MEMS gyroscopic sensors are:  Image stabilization for handheld video. by oscillating a proof mass that is suspended with elastic springs that restrict the mass to move in two degrees of freedom (DOF). such as.2 Operating Principles The majority of MEMS vibratory gyroscopes use the Coriolis force to detect the angular velocity of a rotating body.1) Where 𝑎𝑐 is the Coriolis acceleration. The angular velocity 𝑤 is the characteristic of the rotating body that we intend to indirectly measure. 1. toys and hobbyist applications The application for micro gyroscopic sensors is large and will continue to grow with further research into making them a more enabling technology. The cross product of the velocity and angular velocity gives a resultant acceleration in the orthogonal direction. (1. The Coriolis force is a fictitious force that cannot be observed in the inertial frame. DSLR and point-and-shoot cameras  Computer and video game input devices  Biomechanical research and monitoring. If the velocity and angular velocity of the body act in the same direction then there is no resultant Coriolis acceleration. which the gyroscope is attached to.2. as such an artificial velocity must be imposed on the microgyrscope’s proof mass in order to create a resultant Coriolis force that can be directly measured. human motion tracking  Personal electronics. The Coriolis acceleration is given by the formula: 𝑎𝑐 = 2𝑣 × 𝑤. higher accuracy. 𝑣 is the velocity of the proof mass in the accelerating frame of reference and 𝑤 is the angular velocity of the rotating body.

an oscillating velocity can be used instead. Figure 1-1-1 . The Coriolis force will also be oscillatory in nature as a result. is mounted on some hypothetical rigid body with unknown dimensions and a centre of mass.impose a constant velocity on the proof mass since it would soon leave the confines of the system. Given a hypothetical gyroscope such as in figure 1-1. 1. any displacement in the driving direction will change the length of the springs in the sensing direction and vice-versa. which will affect the underlying forces. G. The process for designing a mechanical system that has the same or similar driving and sensing resonant frequencies is called mode matching and is an important constraint for micro-gyroscopes. M. and thus the forcing frequency of the gyroscope will ideally coincide with the sensing resonant frequency. ideally the frequency of the imposed proof mass velocity must match the natural frequency in which the proof mass will readily oscillate in the sensing direction. 4 . In other words.3 Gyroscope Kinematics The MEMS gyroscope. If the proof mass is not excited at the correct frequency then the displacement in the sensing direction will be significantly reduced. Another important constraint in the design of vibratory gyroscopes is the decoupling frame.Hypothetical vibratory gyroscope. with an angular velocity of w as shown in figure 1-2 below.2. A decoupling frame will ideally mechanically isolate the driving and sensing modes.

By mounting the gyroscope close to the centre of mass of the rigid body then the Euler acceleration becomes negligible and the acceleration acting on the proof mass in the sensing direction becomes: 5 .2) 𝑣𝑀/𝐺 = 𝑥 𝑖 + 𝑤𝑥𝑗 (1. the driving direction corresponds to the 𝑖 direction. Both these accelerations act in the driving direction. In the below case. 𝑗. these accelerations act in the sensing direction. Figure 1-1-2 .Kinematics of a gyroscopic sensor system The kinematic formulas arising from this system are given below: 𝑃𝑀/𝐺 = 𝑥𝑖 (1. The position vector 𝑃𝑀/𝐺 locates the centre of the gyroscopes proof mass to the centre of mass of the rigid body. The direction vectors.3) 𝑎𝑀/𝐺 = 𝑥 − 𝑤 2 𝑥 𝑖 + (2𝑤𝑥 + 𝑤 𝑥)𝑗 (1. 𝑘. 𝑖.An inertial frame of reference is used in this investigation and thus the coordinate system rotates with the rigid body.4) The 𝑥 term in the right side of equation (1.4) is an acceleration imposed on the system by the forced vibration on the proof mass. rotate with the center of mass of the object. The Euler acceleration disappears when the angular velocity of the body does not change. the sensing direction is in the 𝑗 direction and the input rotation is about the 𝑘 direction. The 2𝑤𝑥 and 𝑤 𝑥 terms are the Coriolis and Euler accelerations respectively. the −𝑤 2 𝑥 term is the centrifugal force.

f is the frequency.5) The resultant force is proportional to the angular velocity of the rigid body when the velocity of the proof mass in the driving direction is constant. which is a desirable characteristic for a sensor. 6 . This will. However the velocity of the proof mass in the driving direction is not constant and is in the form: 𝑥 = 𝐴(2𝜋𝑓)𝑐𝑜𝑠(2𝜋𝑓𝑡 + 𝜑) (1.6) where A is the amplitude. This can be corrected by measuring the acceleration separately and adjusting the output signal accordingly. produce an oscillating force and oscillating output signal. 𝑎𝑐 = 2𝑤𝑥 𝑗 (1. in turn. A potential source of error is an outside acceleration acting on the rigid body in the sensing direction causing a resultant force on the proof mass. The velocity of the proof mass in the driving direction is controlled thus the angular velocity can be deduced from measuring the force in the sensing direction. t is the time and φ is the phase angle.

4 Performance Characterization The quality of all gyroscopic sensors can be measured by their performance of each of the below factors:  Scale Factor – The proportional constant of the change of output to a change in input. Some common secondary inputs include temperature.2. All sensitivities must be compared to the operating standard by which the original scale factor was determined. pressure and humidity. therefore the resolution can be determined by measuring the standard deviation of the white noise. 7 .  Sensitivity – As with other sensors. The drift rate is the peak-to-peak value of this function and is measured in o/s. The units of sensitivity are different depending on which undesirable input is considered. The white noise of the device is what limits the resolution. The sensitivity of the output to a secondary input is the ratio of change in the proportional scale factor to a change in the secondary input.1. The unit used is (°/s)/Hz.  Drift Rate – The output of a gyroscope is subject to a slowly changing random function that is independent of the input angular rate.  Resolution – The minimum change in input that will result in a change in output. The scale factor can be found by fitting a straight line to the input and output data using the method of least squares. the output signal can be affected by environmental conditions or other undesirable inputs. The output signal is measured in Volts so the unit of the constant is V/(°/s).

05. The following formula can be used to determine the bandwidth of the gyroscope. The quality factor describes the relationship between the mass-spring and damping coefficients. Increasing the damping in the system will subsequently reduce the amplitude but extend the frequency range of the mode. while a low quality factor represents a small amplitude with a shallow drop in amplitude. A damped mass-spring system that is being vibrated at its natural frequency will experience greater amplitude compared to a similar system that is not. 8 .8) where ∆𝑓 is the range of frequencies and 𝑓0 is the natural frequency. the damping coefficient must be less than 0. (1. ∆𝑓 = 𝑓0 𝑄 . The zero rate output (ZRO) can be found by determining the mean of the output signal in the absence of an input rotation and is measured in Volts. This formula may only be used with a ζ < 0. Similarly to the quality factor calculation. 1 2 𝑄= (1. A high quality factor represents a large amplitude at the natural frequency with a sharp drop in amplitude as the forcing frequency changes from the resonant value. Zero Rate Output – Any inherent dissymmetry or disparity in a micro-gyroscope may result in an offset to the output.  Quality Factor – Mechanical. damped mass-spring systems respond to all forcing frequencies but can have varying amplitudes depending on the natural frequencies of the system.7.  Bandwidth – The bandwidth of the gyroscope is the range of frequencies where the displacement of the proof mass is equal to at least half of the resonant peak value.05. The quality factor can be determined theoretically using formula 1.7) 2𝜁 where Q is the quality factor and ζ is the damping coefficient.

the voltage required is too great to produce a significant force due to the large distance between the two electrode plates.1. 9 . Parallel plates provide a large facing area which enable large force but only for small motions due to rapid change of force when the gap between the two plates varies. there are two sets of fingers or plates. and comb drive actuator shown figure 1. However. The most common micro-actuators are thermal and electrostatic but other types include magnetic and piezoelectric actuators. 𝜀𝐴 𝑉 2 𝐹𝑒 = 2𝑑 2 (1. The dramatic force change causes instability in the actuator. which is connected to flexures and is free to move. In macro scale systems. the distance between the two electrode plates is relatively small. shown in figure 1. but act in different directions due to different configurations of the plates. in the micro scale. micro- pumps. the actuation method used to drive the proof mass in vibratory micro-gyroscopes has been electrostatic actuation. RF. micro-manipulators and inertial sensors. ɛ is the permittivity of air. assuming that the effect of the electrical field on the edges of the plates is negligible.4. one is anchored to the substrate and faces the other.1 Electrostatic Actuation Previously. Both actuators operate by the same electrostatic forces. In parallel plate actuators.9) where Fe is the force from the electric field.3.3. There are two popular MEMS electrostatic actuators. In most macro scale applications electrostatic forces are not used as a method of actuation since the electrostatic force arising from the Coulomb charging of parallel plates decreases with the square of distance. A is the area of the parallel electrode plates. V is the charge of the parallel plates and d is the distance between the plates. 1. the parallel plate actuator.3 MEMS Actuators MEMS actuators are required for a variety of applications including ink-jet printers.

a) Schematic illustration of the general structure of parallel plate actuator b) Micrograph of the prototype Figure 1-1-3 . They are capable of larger displacements compared to parallel plate actuators. Comb drive actuators operate perpendicular to the working plane and use symmetrical fingers with repulsive electrostatic forces.Therefore.3 10 . it is only suitable for applications requiring small motion with large force.Parallel plate actuator.

Indirect heating does 11 . their thickness and width are far smaller compared to their length. Thermal expansion is often considered in the longitude direction since most of MEMS devices are thin and narrow. However. typically 30V to 200V range56 and hence are difficult to integrate with other CMOS (Complimentary-Metal-Oxide-Silicon) circuits.Micrograph of an in-plane Comb Drive Actuator. Thermal actuators can be heated by either direct or indirect heating. 1. Direct heating increases the temperature of the actuator by inducing a current through the volume of the expanding elements where heat is generated through the materials resistance to the current. b) Close-up of the fingers Figure 1-1-4 . This physical process is called Joule heating and is more efficient in comparison to indirect heating. the voltages are relatively high for most applications including vibratory gyroscopes.a) Image of an in-plane comb drive actuator.3. The volume of a material expands as the temperature increases and a force is created as a result if a mechanical constrain is applied to the structure.2 Thermal Actuation Thermal actuators use thermal expansion to generate forces and displacement.4 The advantage of using electrostatic actuation methods is that they require very little power because very little current is required to charge the plates due to their low specific electrical capacity. which is facilitated by asymmetrical geometry or the use of materials with different thermal expansion properties.

11) 𝐴 Where ρ is the material resistivity. Joule heating is given by the formula: 𝑄𝐺𝐸𝑁 = 𝐼 2 ∙ 𝑅 ∙ 𝑡 (1. The material resistivity is dependent on the temperature and changes during operation of the actuator.12) Where m is the mass of the structure in kilograms.10) Where QGEN is the energy is joules. l is the length and A is the cross-sectional area of the structure. I is the current flowing through the structure in amps. c is the specific heat capacity of the material in joules per kilogram-kelvin and T is the temperature in kelvin.not generate heat within the structure but instead uses a proxy source to heat it using conduction. This allows for greater flexibility in the device design and allows for greater control but is less efficient and has a slower response. convection and conduction in joules. t is the amount of time in seconds and R is the resistance of the structure in ohms that is characterized by the formula below: 𝜌𝑙 𝑅= (1. convection and conduction so the temperature of the device can be found by accounting for these terms: 𝑄𝐺𝐸𝑁 − 𝑄𝑅𝐴𝐷 − 𝑄𝐶𝑂𝑁𝑉 − 𝑄𝐶𝑂𝑁𝐷 = 𝑚 ∙ 𝑐 ∙ ∆𝑇 (1. radiation or convection. Conduction is the dominant heat transfer mode where most energy is transferred to the substrate which acts as a large heat sink. thus they are typically ignored. Radiation is not very significant at the temperatures that thermal actuators operate in and convection is negligible in the micro-scale. Some of the energy generated from Joule heating is dissipated into the surroundings through radiation. 12 . Using direct heating methods. The Q terms represent the energy generated and the energy lost from radiation.

16) Thermal actuators are compact. The three most common thermal actuators are bimorph.14) The force can be found from the stress. and therefore the displacement of the actuator can be found from the temperature according to the thermal expansion strain formula: 𝜀 = 𝛽∆𝑇 = ∆𝐿 𝐿 (1. As the beam heats up. usually containing a thin and wide arm to create displacement. the Young’s modulus of the material. MEMS bimorph actuators have fast response times and can vibrate at relatively high frequencies due to its low specific thermal capacity and light weight. Several “V” shaped beams are placed in parallel and fixed at both ends. The point of the chevron actuator moves 13 .13) Where ß is the material specific. The actuator bends towards the thick arm due to the difference in arm lengths. linear thermal expansion coefficient defined as: ∆𝐿 𝐿 𝛽= ∆𝑇 (1. E. ɛ. The resistance in the thin arm is greater than the thick arm due to its smaller cross-section causing it to generate more heat and expand more than the thick arm. require low voltages and generate large displacements in comparison to electrostatic actuators.Strain. σ. the strain and the cross-sectional area: 𝜍 = 𝐸𝜀 (1. the beam bends toward the direction of the lower thermal expansion coefficient layer. hot/cold arm and chevron. reliable. Chevron actuators also use device asymmetry to generate displacements. Bimorph actuators are typically in the form of a cantilevered beam made from two separate material layers.15) 𝐹 = 𝐴𝜍 (1. Hot/cold arm actuators use geometrical asymmetry.

MEMS thermal actuators suffer from the same drawbacks but scaling laws dictate that they are negligible in comparison. however the current will also increase and will ultimately require more power compared to electrostatic actuators.adjacent to the beams when the beams expand due to Joule heating. Using a material with a high conductivity will greatly reduce the resistance and the electric potential across the actuator. Chevron actuators generate large forces but are limited in the displacement they can produce. The same techniques are difficult to apply to the micro-scale. The energy required to heat a macro-scale actuator is uneconomical and the heating and cooling response would be slow due to the low surface to volume ratio compared to MEMS sized devices. Alternative methods are required for smaller features and greater resolution. 1. Thermal actuators are also typically restricted to MEMS applications. which also require power to operate.4 Fabrication Fundamentals Traditional machining methods are commonplace and well understood in macro-scale manufacturing settings. The lower voltage requirement of thermal actuators allows them to be easily used with IC`s and need less supporting actuation signal circuitry. Thus. To produce significant travel the actuator must be relatively long. The thermal actuation forcing frequency is also similar to the natural frequencies of MEMS gyroscopes. which would also increase the material cost. Depending on the material and design of a thermal actuator it can use less voltage compared to electrostatic actuators. which also helps reduce the operational power. Lithography and molding methods are well- 14 . it is within reason to expect that a thermal actuator can operate under a lower power requirement compared to a similar electrostatic actuator. Material property discontinuities become significant at sub-millimeter levels as grain boundaries span larger proportions of the width of the tool this are more prone to failure.

A Finite Element Analysis (FEA) is performed on each model to determine the mode shapes of the gyroscopic sensors experimentally. Traditional machining technologies can be used on a wider variety of materials and allow for more control and higher dimensionality structures. Research in this field has led to several promising fabrication technologies such as laser micromachining and micro-electro-discharge machining that are serial processes that allow for better dimensionality. which can be used to assist in the design of future MEMS vibratory gyroscopic sensor designs. The mode shapes are determined experimentally by separate in-plane and out-of-plane tests that produce frequency response data. The parallel characteristic of the majority of lithography techniques make them suited for high-volume production.5 Thesis Outline The objective of the work outlined in this thesis is to theoretically and experimentally determine the modal behavior of three different gyroscopic sensor models and compare the performance of each model. Lithography methods are typically restricted to lower dimensionality structures and the use of specific materials that are sometimes a design constraint. 1. Comparing the simulation and experimental results will improve the theoretical model.understood technologies that can be used for microfabrication applications. The insights gained from these tests will help predict the performance of the sensors. 15 . The information gained by performing these simulations will assist in interpreting the experimental results.

a mask is used to transfer a pattern to a deposited photosensitive material. They are inherently parallel. which is shown in figure 2-1 below. which makes it either susceptible or resistant to the developer solution. by exposing it to radiation. In a positive resist the radiation increases the susceptibility of the resist to the developer. commonly called a resist. In a negative resist the radiation hardens the resist to the developer. Chapter 2 Literature Review 2. Optical lithography is typically used in both processes for patterning. Radiation changes the material properties of the resist. 16 . The two lithography modes that are involved in the Metal MUMPS and Micragem processes.1 Lithography Modes Lithography is a general term that encompasses various geometrical structuring microfabrication techniques. which is useful to increase throughput but are also highly material critical. In optical lithography. which were used to fabricate the micro-gyroscopes. are pattern etching and deposition.

Photolithography process steps7. an additive layer can be deposited to fill in the gaps in the resist and then the resist can be removed to transfer the pattern or an etchant can be used to remove material from a layer beneath the resist. has etching rates that are faster in the vertical direction compared to the horizontal. which produces structures that have a more rounded shape. wet and dry etching. Isotropic etching. etch rate. When used sacrificially. The resist can then be used as part of the geometrical structure or as a sacrificial material that will eventually be completely removed by the end of the process. shown below on the right.1. and material selectivity are all important considerations when choosing which process to use. 2. Anisotropic etching. Anisotropic pyramid shaped structures can be created by etching material between crystallographic planes. shown in figure 2-2 on the left.1 Etching Etching can be separated into two classes. has an etching rate that is equal in both horizontal and vertical directions. 17 . Figure 2-1 . The anisotropy. Undercutting of the mask occurs in both cases and has to be accounted for in the mask design.

Wet chemical etching is typically isotropic but anisotropic structures can be created by taking advantage of crystallographic plane orientations. The feature side walls are almost completely vertical and hence there is no undercutting of the mask. plasma etching is isotropic. The polymer dissolves slowly in the reactive gas leaving it to coat the side-walls of the structure and preventing the reactive ions from etching in the horizontal direction. The anisotropy of side walls can be controlled by adjusting the pressure and energy of the beam. 18 . Sputter etching involves physically bombarding the workpiece with ions to remove the unmasked material. plasma and reactive ion etching (RIE). Figure 2-2 . Preventative measures must be taken when using plasma etching to ensure that chemical by-products do not build up and interfere with the etching process. Alternatively. Reactive ion etching combines sputter and plasma etching by bombarding the exposed material with reactive ions. most dry etching is anisotropic regardless of the presence of crystallographic planes.Anisotropic (left) and isotropic (right) wet etching. sputter. and can be up to 100µm deep. high pressure and reactive gas ions to induce chemical reactions on the surface. In deep reactive ion etching (DRIE) a polymer is deposited on the substrate first and is subsequently removed by the physical bombardment of ions. The structures produced by DRIE are highly anisotropic. 50:1 aspect ratios are possible. In contrast to the other types of dry etching. There are three types of dry etching. Plasma assisted etching uses heat.

Dry etching is expensive compared to wet etching and is subject to worse material selectivities. PVD involves the deposition of a thin film from a vapour to a wafer. which is then deposited on the wafer during evaporative PVD. which makes CVD an enabling technology for MEMS devices. The etch rate for both types is highly dependent on the operational conditions. Using the CVD process. The atoms from the source material are transferred to the wafer by bombarding the source with ions. thermal oxidation. Low pressure and high temperatures are used to vaporize the thin film material.2 Deposition While there are many deposition techniques available. 2. There are several types of CVD that can be used for a variety of materials. the wafer is placed inside a reactor chamber where a chemical reaction occurs between gas molecules. which is then deposited on the wafer. Sputtering uses lower operating temperatures than evaporation but still requires low pressures. The film can be either evaporated or sputtered onto the wafer. Similarly. chemical vapour deposition (CVD). spin coating and electroplating are the most common in thin films. epitaxy.1. physical vapour deposition (PVD). however PVD does not require a chemical reaction for the material deposit to take place. The energy from the collision is enough to eject atoms from the source in the form of vapour. However dry etching allows for greater dimensional control and does not suffer from contamination issues from the etchant such as with wet etching. PVD is cheaper and less dangerous to the wafer compared to CVD but the quality of the thin film is usually worse. The product of the reaction gets deposited on the wafer as a thin film. The desired geometrical structure of MEMS devices will typically determine which etching process must be used. 19 .

The diagram. Epitaxy is similar to CVD but gas or liquid can be used. shows some of the steps involved in using electroplating to fabricate a comb-drive resonator. Spin coating is a useful method for depositing thin polymer films. 20 . The most common application is silicon dioxide. The solvent is evaporated and a thin film remains. oxygen-rich environment to speed up the diffusion of the oxygen atoms in the oxidized material. also referred to as an oxide layer. For example. however the seed layer and epitaxial layer do not necessarily need to be the same material. 2.3 Electroplating Electroplating is a relatively mature fabrication technology in both the micro and macro scales. The substrate acts as a seed layer that defines the epitaxial growth of the layer being deposited on it. IBM first used electroplating as a microfabrication process in the 1960s and the fundamental operating principles have not changed since then. in figure 2-3 below. The material is dissolved in a solvent and then either sprayed or spun onto the wafer. Oxidation time increases with the thickness of the material so there is a constraint on the thickness of the oxide layer.Thermal oxidation is a relatively simple and cheap deposition technique. which is limited to materials that can be oxidized.1. Thick films greater than 100µm with high growth rates are possible using epitaxy. an amorphous silicon wafer will cause an amorphous silicon layer to grow. The process involves using lithography to define a mould and electrodeposition to form the structure. The wafer is placed in a high temperature.

21 .Cross-sectional diagram of an electroplated nickel comb-drive resonator8. is defined using optical lithography and a sacrificial copper layer is grown in a copper pyrophosphate bath. The polyimide layer is used to protect the aluminum from the environment. A gold plating base is added that will be used to grow the subsequent nickel layer from a nickel sulphamate bath. The resonators are released by selectively etching the copper and gold base. labeled PR in the diagram. An SEM image of a released comb- drive finger is shown in figure 2-4 below. The first cross-section shows an oxidized silicon wafer with aluminum electrical interconnects used for ICs and electrodes. The photoresist is stripped and the nickel layer is grown. The positive photoresist mould. Figure 2-3 .

the process involves harmful chemical solutions that can damage preceding layers. Figure 2-4 . Electroplating is commonly used in a process called LIGA that combines X-ray lithography. 22 . However. the dimensional control of LIGA requires highly-focused X-ray beams that are generated by synchrotron radiation. low surface roughness and low residual stress.SEM image of a comb-drive finger fabricated with the electroplating process. However. Recent reports state that LIGA is capable of a lateral resolution better than 0. 1. By taking advantage of economies-of-scale the cost per device can be lowered by using moulding.3mm structural depth. Electroplating requires few process-steps and allows for tall structures with high aspect ratios that are not possible with other microfabrication processes. steep walls up to 100nm per 100 microns and a side-wall surface roughness of 20nm root-mean square depending on the X-ray wavelength used9. The finished product has a high aspect ratio. electroplating and moulding. which makes electroplating an attractive geometrical structuring technique for MEMS devices.5 microns. which restricts the materials that can be used. which are relatively expensive.

The process involves a vacuum chamber. 23 . two electrodes that apply an electric field of 200-1000V across the wafers and heating plates that are used to assist the migration of sodium and oxygen atoms. which causes the two wafers to be electrically attracted to one another. Once a field is applied.4 Anodic Bonding Anodic bonding is a geometrical structuring technique that involves the assembly of a silicon substrate to a borosilicate glass wafer by means of an electric field and thermal bonding.1. with the glass being the cathode and the silicon wafer being the anode. Figure 2-5 – Diagram of anodic bonding between a borosilicate glass and silicon wafer 10. the sodium and oxygen ions migrate towards the electrodes as shown in the diagram below.2. Oxidation at the borosilicate glass-silicon interface bonds the two wafers preventing them from falling apart once the electric field is removed. Very little bulk modification in the silicon wafer is caused by this process. The temperature range is typically controlled to be between 200-450°C.

The process begins with an n-type (100) silicon wafer.1 Metal Multi-User MEMS Process (MUMPS) The metal MUMPS process was designed for general purpose micromachining for MEMS devices. Another 1. A 0. used for anchoring. The structure is released using a 49% HF solution to remove the sacrificial PSG and the oxide layer11. This process was used to fabricate both an electro-thermally driven and electrostatically driven gyroscope.35μm silicon nitride layer is deposited.7μm polysilicon layer used for resistors and additional mechanical structures. The polysilicon layer is lithographically patterned using RIE and a second 0. 500nm of Copper and 50nm of Titanium is deposited as a plating base layer and nickel is electroplated to a thickness of 20μm with a 0. Both nitride layers are then patterned by RIE as well. The thickness of each layer is exaggerated so that it can be seen.1μm PSG layer is added and then annealed at 1050˚C for 1 hour. Once the stencil is removed a 1. The cross-sectional diagrams below outline the Metal MUMPS steps as it applies to the microgyrscope designs.35μm silicon nitride layer.2 Microfabrication Processes 2. 24 .5μm gold layer electroplated on top.5μm sacrificial phosphosilicate glass (PSG) layer. which is then patterned using wet chemical etching. is deposited as well as a 0.2. An electroplated nickel layer is used as the primary structural material and also for electrical interconnects. A 2μm oxide layer is then grown on the wafer followed by a 0.2. used for electrical isolation. The oxide is etched away and a metal layer. A photoresist stencil was used to pattern the nickel and give it an anisotropic structure.3μm gold sidewall is electroplated. is deposited and patterned using lift-off.

In c) the polysilicon is patterned. The cross-section d) in figure 2-6 shows the second.Cross-sectional view of an electrostatic gyroscope fabricated with the Metal MUMPS process. In e). patterned PSG layer after the annealing process. PSG. 25 . silicon nitride and polysilicon layers are then added in b). However. the plating base layer is added and the nickel is grown by electroplating in the gaps between the photoresist. which is then patterned as well. The gaps in the PSG layer are used to affix the subsequent nickel layer to the substrate. a) b) c) d) e) f) Figure 2-6 . Figure 2-6 a) shows the plain silicon wafer. Step f) shows the released structure after the sacrificial base layers have been etched away. followed by an additional nitride layer. the oxide layer cannot be seen because it was not utilized in this case. A 3 dimensional view of the final product can be seen in the figure below. The oxide.

3D view of an electrostatic gyroscope using the Metal MUMPS fabrication process. The process begins with a 500µm glass wafer that is then selectively etched using to add gaps or cavities to the structure.2. An additional SOI layer is then anodically bonded to the glass and the silicon handle and buried oxide is etched away to leave the 10µm single crystal silicon layer (SCSi) behind. Metal electrodes and bonding pads consisting of 500Å Titanium- Tungsten and 2000Å gold are added using surface patterning.2 Micragem Silicon-On-Insulator (SOI) Micragem is a microfabrication process developed by Micralyne Inc. and CMC Microsystems for MEMS applications. The final step is to etch the SCSi layer using DRIE12. An additional metal layer can be deposited on top of the SCSi layer at this point and patterned using lithography. 2. which makes the process comparatively versatile and increases the design possibilities for micro-gyroscopes. Decoupling Frame Proof Mass Driving Frame Sensing Fingers Parallel Plate Actuators Figure 2-7 . 26 . The technology allows for fully suspended metal electrodes and three different gap depths.

The next diagram. d).Cross-sectional views of an electrostatic gyroscope using the Micragem process.These steps are shown visually below where the Micragem process was used to fabricate an electrostatically driven micro-gyroscope. shows the SOI layer being joined to the original substrate using anodic bonding. The bonding pads are added in the c). The thicknesses have been scaled to show off the significant features of each layer. The oxide and silicon handle are removed by etching in e) and then the SCSi layer is patterned in the last phase. 27 . The cross-sectional diagram in figure 2-8 a) shows the bare glass substrate. which will be used to electrostatically actuate the subsequent SCSi layer. The end result can be seen in the figure below. a) b) c) d) e) f) Figure 2-8 . The surface is etched to add gaps and cavities in the structure in b).

Driving Capacitive Plate Proof Mass (Underneath Structure) Driving Frame Contact Pads Figure 2-9 . The device was assembled by hand and contained a silicon member and a separate glass slide containing the electrodes. which influenced the invention of the first micro-gyroscope created in 1988 by Draper Labs and was presented at the Navigation and Control Conference14. 2. Their research was instrumental in verifying the micro-gyroscope concept and establishing a model for which proceeding gyroscopes would be compared.3 Previous Micro-gyroscope Research The first MEMS inertial sensor was an SOI accelerometer developed by Roylance et al. The footprint of the outer gimbal was 350x500µm2. The operational characteristics of the device are graphically shown in figure 2-10 below. Draper Labs published a paper on another double-gimbal micro-gyroscope supported by torsional flexures with an on-board buffer circuit to reduce noise and made from single-crystal silicon.3D view of an electrostatic gyroscope using the Micragem process. that a larger gyroscope would elicit greater performance. This development pioneered MEMS inertial sensors. it was hypothesized by Boxenhorn et al. In 1991. in 1979 for applications that required small transducers13. The gyroscope was designed to detect the 28 .

z-axis. Capacitive sensors were used to detect the resultant motion of the inner frame. at a footprint of 400x800µm2. The device was slightly larger. an oscillating torque was applied in the drive axis electrostatically by silicon p-type electrodes on the ends of the device.angular velocity of the out-of-plane. The Coriolis force caused the inner frame to oscillate in the output axis direction at the same frequency as the driving frequency and at an amplitude that was proportional to the magnitude of the angular rate of the input. The substrate was a p-type silicon wafer that was selectively doped to make electrodes. the device would be driven at the resonant frequencies of the sensing gimbal to maximize the resultant amplitude. When operated within a vacuum at 3kHz a 4o/s/√Hz resolution was demonstrated.A novel double-gimbal micro-gyroscope by Draper Labs The first lateral-axis gyroscope was developed Tanaka et al. component of the rotating system that it was mounted to. A 1µm sacrificial PSG layer was deposited to separate the subsequent 5µm polysilicon layer from the substrate. compared to Draper Labs double-gimbal gyroscope. Figure 2-10 . in 199515. Oxide and nitride layers were added for insulation. To detect this angular rate. 29 . Ideally.

Suspended Beams Suspended Beams Figure 2-11 . Electrostatic comb actuators and sensors were used to drive and detect the motion of the proof mass respectively. which was proportional to the angular velocity.800 and 16. An angular velocity applied in the y-direction causes the proof mass to vibrate out-of-plane in the z-direction due to the Coriolis force. A piezoelectric actuator was mounted to the gyroscope to detect mechanical Q factors at various pressures. the proof mass was suspended by 4 beams and was driven in the x-direction by an AC voltage. coupled micromachined vibratory gyroscope. At 1Pa Q factors of 2. 30 . The technique involves modifying the stiffness in the supporting springs. which are relatively sharp peaks.A planar. Ion milling was used to decrease the height of the spring. In figure 2-11 above. also conceived of a technique to adjust the driving and sensing mode frequencies so that they are close to each other in 1995. and at an atmospheric pressure below 0.000 were found for the driving and sensing modes respectively. and RIE was used to modify the width of the spring reducing the driving mode resonant frequency.475kHz. Tanaka et al. Electrodes beneath the mass detected the displacement in the z-direction. therefore reducing the sensing mode resonant frequency.1Pa a resolution of 7o/s/√Hz was found. 12. Using a 5V DC bias and 10V peak-to-peak AC driving voltage operating at the driving mode resonant frequency of the device.

which is depicted in the figure below. Four 180µm long symmetrically placed beams were used to suspend the mass 1. Sensing was performed by four electrodes beneath the proof mass. Using an open-loop driving signal the team was able to demonstrate resolutions of 0. at the frequency of the drive oscillation. Their dual input axis device was manufactured using Analog Devices Inc.6µm above the substrate. The proof mass of the gyroscope was a 2µm thick circular disk with 150µm radius. Any rotation about the x-axis would induce an oscillating torque. oscillate the inertial element orthogonally to the substrate.’s paper in 1997 only single input axis gyroscopes had been researched16. During operation.6µm above the substrate and to oscillate the inertial element orthogonally to the substrate. which was then detected electrostatically using the electrodes underneath.a) Conceptial diagram of a dual input axis gyroscope. Drive combs. Four 180µm long symmetrically placed beams were used to suspend the mass 1. b) A micrograph of the proposed gyroscope. The proof mass of the gyroscope is a 2µm thick circular disk with 150µm radius. a) b) Figure 2-12 . proprietary surface micromachining process. the inertial mass was oscillated about the z-axis.Until Juneau et al. The functional concept requires symmetry in two planes normal to the substrate. The torque tilts the proof mass. on the proof mass about the y-axis and vice versa due to the induced Coriolis force. shown in figure 2-12 b) above.24 o/s/√Hz by matching the 31 .

Mochida et al. predicted that using closed-loop force feedback control would reduce the cross axis dependency and increase device performance. The resultant structure is shown pictorially in figure 2-13 below. and investigated the cross axis sensitivity and noise floor of both to determine the effect of coupling. Juneau et al.driving and sensing modes but with high cross axis sensitivity.MEMS gyroscope utilizing independent drive and sense beams for mode decoupling. were able to reduce mode coupling mechanically by using separate beams for the driving and sensing modes in 199917. The glass and SOI layers were then anodically bonded together and the silicon handle was etched. 32 . Sensing axis coupling was reduced when the driving frequency was unmatched with respect to the sensing natural frequency but increases the noise floor. A glass substrate was then wet etched and gold and chromium layers were patterned as electrodes. The size of the resonator was 800x1200x50µm3. The mechanical structure was defined by etching the top silicon layer using RIE. In their study they fabricated a similar device to Tanaka et al. The device was fabricated from an SOI substrate. Figure 2-13 . The process is similar to the standardized Micragem fabrication process.

’s design. Overall. Figure 2-14 . Previously. The frame was supported by four beams similarly to Tanaka et al. The noise was greater for Mochida’s design. Using laser interferometry the cross axis coupling was experimentally measured and compared. were able to develop a micro-gyroscope that compensated for fluctuations in the 33 . the reduced coupling improved the resolution by 100 times that of Tanaka’s device with a resolution of 0.07 o/s/√Hz at 10Hz bandwidth while operating at an environmental pressure of 100Pa.’s device had half of the x-z amplitude ratio compared to Tanaka et al. which is likely due to less mechanical stability due to the increased DOF of the system. the deflection of the inner frame does not change with a deflection in the outer beams and vice versa. this has been compensated by post-processing of on-chip electronics.’s device for all tested frequencies. Acar et al. Ideally.Drive mode beam deflection (left) and sensing mode deflection (right). The path of the proof mass in the x-z plane when driven electrostatically was elliptical in shape for both gyroscopes. however some coupling was observed in the FEM model but very weak in comparison. One of the biggest challenges in the previous micro-gyroscope designs were the variability in output due to residual manufacturing stresses and environmental temperature and pressure fluctuations. Figure 2-14 below shows the intended deflection of the beams in their respective modes of operation.The decoupling frame contains four springs allowing the proof mass to move in the z-direction. However Mochida et al.

Figure 2-15 .environment and residual stresses mechanically.SEM image of a robust 3 DOF mechanically decoupled gyroscope. The gyroscope uses two DOF for the sensing mode and one DOF for driving and incorporates a decoupling frame. which helped reduce the complexity and increase the robustness of the system18. An SEM image of the device is shown in figure 2-15 below. The mass m1 is free to move in the driving and sensing direction while mass m2 is constrained in the driving direction with respect to mass m1. This displacement was detected using parallel plate capacitive sensors. 34 . The footprint of the device is 4x4mm2. This produces two separate response peaks since the two masses have separate resonant frequencies as shown in figure 2-16 below. The two masses oscillate together in the drive direction using comb-drive actuators and their response in the sensing direction is the sum of the response of both masses in the sensing direction.

64o/s/√Hz at 50Hz bandwidth. Temperature insensitivity was also proven to be 12. The masses were driven between the two sensing peaks where the response amplitude is flat.2 times less compared to a conventional one DOF micro-gyroscope. The device was installed onto a rate-table and excited using a 25V DC bias with a 3V AC drive signal and operating at atmospheric pressure. In the given frequency range the gain is less sensitive to changes in the natural frequencies of the system making it more resilient to fabrication imperfections and environmental changes. 35 . Figure 2-16 – Frequency response curves of a 3 DOF gyroscope. It was found that the flat region between the peaks remained relatively constant despite the variations in temperature and pressure while the amplitude of the peaks decreased with variations in pressure and shifted with changes in temperature. The device was tested under a variety of environmental conditions. The resolution of the device was then determined to be 0.

A 3D modal analysis of each gyroscope was performed to predict the experimental modal frequencies. free-floating mechanical structure is shown in the figure below. Chapter 3 MEMS Gyroscope Theoretical Modal Analysis In this section the simulated models of three comparable MEMS gyroscopes are developed using COMSOL Multiphysics finite element analysis software. 3. Figure 3-1 .1 Thermally-Actuated Gyroscopic Sensor Analytical Model A visual representation of the fixed anchors and suspended. A simplified analytical model of the thermally-actuated gyroscopic sensor is also presented.Thermally-actuated gyroscope FEA boundary condition definition diagram. 36 .

1. The proof mass is attached to the decoupling frame using four triple-folded flexures. and act as rigid beams in the other directions. 𝜑. The DoF are shown in a diagram of the proof mass below. from the motion of the driving frame. 37 . ky Decoupling ky ky Frame ky Mass kx kx Proof kx Mass kx ky ky ky ky kchev Figure 3-2 .1 Analytical Model Assumptions and Constraints An analytical model of the thermally-actuated gyroscopic sensor was developed to support the simulated and experimental results. The chevron and double- folded flexures are modelled similarly. with mass M. 3.Thermally-actuated gyroscopic sensor in-plane simplified model. which ideally act as linear springs in the sensing direction. This restricts the system such that there are only 5 DoF present. The decoupling frame. The thermal actuator is used to drive the decoupling frame that is suspended above the substrate using eight double-folded flexures that are anchored to the substrate. x. z. of mass m. is used to mechanically decouple the proof mass. The picture of the gyroscopic sensor in figure 3-1 can be simplified into the following diagram. y. θ.The device itself has an overall footprint of 6mm2 and is 20µm thick. To eliminate any coupling of the proof mass and decoupling frame in the model it must be assumed that ky<<kx and ky<<kz for the double-folded flexures attached to the decoupling frame and kx << ky for the triple-folded flexures attached to the proof mass. the x- direction in this case.

1) 𝑑𝑥 2 Where: 38 . kframe. x. which is treated as a spring. Where di and dj are the distances between the triple-folded flexures.2 Spring Modelling The displacement caused by the flexures can be modelled by simplifying each fold in the flexure to be an independent cantilevered beam. where E is the Young’s modulus of the material. It is assumed that the frame is rigid except for the thin section of the decoupling frame in the z-direction. 𝑑2 𝐸𝐼 ζ 𝑥 = −𝑀(𝑥) (3. 3. ζ . +y + 𝜑 z di + Proof 𝜃 dj x Mass k3-fold k3-fold k3-fold k3-fold kframe kframe Figure 3-3 . F. The Euler-Bernoulli relation can be used to find the displacement of the cantilevered section.Out-of-plane thermally-actuated gyroscopic sensor model.1. I is the moment of inertia of a beam and M is the moment as a function of the displacement from the wall. L. F x L Figure 3-4 . can then be summed to receive the total displacement of the flexure. Several other assumptions were made to develop the analytical model. applied to the flexure fold section of length.Folded flexure simplified model. The diagram below shows a force. The displacement of each fold.

The lengths of each flexure are identified in the figure below. 𝑤𝑡 3 𝑀 𝑥 = −𝐹(𝐿 − 𝑥) . Lx Ly Lc Figure 3-5 . 𝐼 = 12 The boundary conditions that were used were: 𝑑 ζ 0 = 0 .2) 3𝐸𝐼 The displacements of each fold can then be summed to receive a total displacement of the flexure. the total deflection of the flexure in the x-direction will be: 3𝐹𝐿3𝑥 12𝐹𝐿3𝑥 𝜁𝑥 = = 3𝐸𝐼 𝐸𝑡𝑤 3 39 .ζ 0 = 0 𝑑𝑥 Solving for this provides the following equation for the displacement at the end of the cantilevered beam: 𝐹𝐿3 ζ 𝐿 = (3. 19 Thus. there are three long sections of flexure length Lx that are orthogonal to the intended spring direction and four short sections of length Lt that are parallel to the spring direction.Analytical modelling of the flexures. For a triple-folded flexure.

which acts a distance. 40 . F.6) 12𝐿3𝑥 + 16𝐿3𝑡 The stiffness of the decoupling frame is predicted to be significant as well. the double folded flexure spring constants in the y-direction are: 𝐸𝑡𝑤 3 𝐸𝑡𝑤 3 𝑘𝑦1 = . the folded sections also act as independent cantilevers. from the fixed ends as shown in the figure below. The deflection caused by the decoupling frame can be evaluated by assuming that the combined stiffness of the double- folded flexures and chevron actuator that are attached to the decoupling frame are negligible in comparison. 𝑘 𝑦2 = (3. therefore the spring constants can be calculated similarly below. Therefore.Where w is the width of the flexure and t is the thickness. d. the spring constant of the triple-folded flexure is: 𝐸𝑡𝑤 3 𝑘𝑥 = (3. However. 3𝐹𝐿3𝑥 4𝐹𝐿3𝑡 12𝐿3𝑥 + 16𝐿3𝑡 𝜁3−𝑓𝑜𝑙𝑑 = + = 𝐹 𝑤𝑡 3 𝑤𝑡 3 𝐸𝑤𝑡 3 3𝐸( 12 ) 3𝐸( 12 ) 𝐸𝑤𝑡 3 𝑘3−𝑓𝑜𝑙𝑑 = (3. caused by the proof mass acts on the decoupling frame of length L through the triple-folded flexures.4) 8𝐿3𝑦1 8𝐿3𝑦1 Where Ly1 and Ly2 are the length of the double folded flexures. The thin section of the decoupling frame that is connected to the triple-folded flexures that support the proof mass can then be assumed to have fixed ends. A force.5) L3c The three folded flexures also act as springs in the out-of-plane direction. The chevron actuator also acts as a spring and is calculated using the following formula19: 16Etw 3 k chev = M (3.3) 12𝐿3𝑥 Similarly.

Equation 3.1 can be used again to model the above diagram with the following boundary conditions: 𝑑 𝐿 𝜁 = 0 .𝑧 0 = 0 𝑑𝑥 2 The formula of the moment is: 𝑀 = −𝐹 𝑥 + 𝐹 𝑥 − 𝑑 + 𝐹 𝑥 − 𝐿 − 𝑑 Solving for the displacement at distance d from the fixed ends yields: 𝑑𝐿2 (𝐿 − 𝑑)3 3𝑑𝐿2 −𝑑3 − 8 + 6 −12𝑑3 − + 2(𝐿 − 𝑑)3 𝜁 𝑑 =𝐹 =𝐹 2 𝑓𝑟𝑎𝑚𝑒 𝐸𝐼 𝐸𝑤𝑡 3 This results in a spring constant for the decoupling frame of: 𝐸𝑤𝑡 3 𝑘𝑓𝑟𝑎𝑚𝑒 = (3.Decoupling frame forces in 𝜑 mode shape. F F d d L x Figure 3-6 .Decoupling frame simplified model.7) 3𝑑𝐿2 −12𝑑3 − 2 + 2(𝐿 − 𝑑)3 The forces acting on the decoupling frame for the 𝜑 mode shape are similar to figure 3-6 except the forces are acting in opposing directions as shown below: F F d d L x Figure 3-7 . 41 .

is: 42 .Which then gives the spring constant for the decoupling frame about the i-axis: 𝐸𝑤𝑡 3 𝑘𝜑−𝑓𝑟𝑎𝑚𝑒 = (3.8) −2𝑑3 𝐿 − 4𝑑4 − 12𝑑2 𝐿2 + 18𝑑𝐿3 + 4𝐿4 𝐿 The independent springs are then summed in parallel to give the equivalent stiffness in the x. Numerical solutions are given below: 𝐸𝑡𝑤 3 219𝐺𝑃𝑎 (20𝜇𝑚)(13𝜇𝑚)3 𝑘𝑖 = 4𝑘𝑥 = 4 =4 = 182. and z directions defined as ki. kk’.50𝑁/𝑚 12𝐿3𝑥 12(260𝜇𝑚)3 𝐸𝑡𝑤 3 𝐸𝑡𝑤 3 16Etw 3 𝑘𝑗 = 4𝑘𝑦1 + 4𝑘𝑦2 + 6𝑘𝑐𝑕𝑒𝑣 = 4 + 4 + 6 8𝐿3𝑦1 8𝐿3𝑦2 L3c 219𝐺𝑃𝑎 (20𝜇𝑚)(8𝜇𝑚)3 219𝐺𝑃𝑎 (20𝜇𝑚)(8𝜇𝑚)3 = + 2(362𝜇𝑚)3 2(420𝜇𝑚)3 16 219𝐺𝑃𝑎 (20𝜇𝑚)(8𝜇𝑚)3 +6 = 377.24𝑁/𝑚 (860𝜇𝑚)3 −1 −1 1 1 1 1 𝑘𝑘 = + = + 4𝑘3−𝑓𝑜𝑙𝑑 2𝑘𝑓𝑟𝑎𝑚𝑒 𝐸𝑤𝑡 3 4 𝐸𝑤𝑡 3 12𝐿3𝑥 + 16𝐿3𝑡 2 3𝑑𝐿2 −12𝑑3 − 2 + 2(𝐿 − 𝑑)3 1 = 3 219𝐺𝑃𝑎 13𝜇𝑚 20𝜇𝑚 4 12 260𝜇𝑚 3 + 16 20𝜇𝑚 3 −1 1 + 219𝐺𝑃𝑎 (20𝜇𝑚)(20𝜇𝑚)3 2 3(400𝜇𝑚)(1400𝜇𝑚)2 −12(400𝜇𝑚)3 − 2 + 2(1400𝜇𝑚 − 400𝜇𝑚)3 = 107.64𝑁/𝑚 The spring constant for the 𝜑 mode. y. kj and kk respectively.

The mass and mass moments of inertia are given below.3 Analytical Resonant Frequencies Calculation The Lagrangian equations of motion were developed for the system given the degrees of freedom: x.8) 2 2 2 1 1 1 1 𝑇 = 𝑚 𝑥 2 + 𝑦 2 + 𝑧 2 + 𝑀𝑦 2 + 𝐼𝑗 𝜃 2 + 𝐼𝑖 𝜑2 (3.19 × 10−14 𝑘𝑔 ∙ 𝑚2 12 Assuming small deflections about its neutral position.46𝑁/𝑚 1 1 + 4𝑘3−𝑓𝑜𝑙𝑑 2𝑘𝜑−𝑓𝑟𝑎𝑚𝑒 3. z.8012 × 10−7 𝑘𝑔 1 𝐼= 𝑚 𝑏 2 + 𝑕2 12 1 𝐼𝑖 = 1.77 × 10−15 𝑘𝑔 ∙ 𝑚2 12 1 𝐼𝑗 = 1.9) 2 2 2 2 Where x. 1 𝑘𝑘 ′ = = 233. 𝜑.9224 × 10−7 𝑘𝑔 𝑀 = 1. 𝑚 = 1. g is the gravitational acceleration and Ij and Ii are the mass moments of inertia of the proof mass. y and z are the displacements of the proof mass and decoupling frame. The equations for the potential and kinetic energy of the system are shown below. 1 1 1 𝑈 = 𝑘𝑖 𝑥 2 + 𝑘𝑗 𝑦 2 + 𝑘𝑘 𝑧 2 + 𝑑𝑖2 𝑠𝑖𝑛2 𝜃 + 𝑑𝑗2 𝑠𝑖𝑛2 𝜑 − 𝑚𝑔𝑧 (3. θ.1. y.9224 × 10−7 𝑘𝑔 600𝜇𝑚2 + 20𝜇𝑚2 = 5. the above equations are simplified to: 𝑥: 𝑚𝑥 + 𝑘𝑖 𝑥 = 0 𝑦: 𝑚𝑦 + 𝑘𝑗 𝑦 = 0 𝑧: 𝑚𝑧 + 𝑘𝑘 𝑧 = 0 𝜃: 𝐼𝑗 𝜃 + 𝑑𝑖2 𝑘𝑘 𝜃 = 0 𝜑: 𝐼𝑖 𝜑 + 𝑑𝑗2 𝑘𝑘 ′𝜑 = 0 The below solution is assumed for the system: 43 .9224 × 10−7 𝑘𝑔 1800𝜇𝑚2 + 20𝜇𝑚2 = 5.

{𝑥 𝑡 } = 𝑋 𝑒 𝜆𝑡 The natural frequencies of the system. 6. z-direction. A 3D tetrahedral element. where dark red and dark blue indicate large displacements from the neutral axis and teal indicates negligible displacement. Colouring was also used as an indicator for the out-of-plane. can be solved in the following matrix. 3. size-varying mesh was used in each case.066kHz.766kHz. The pre- packaged MEMS material constants for nickel and SCSi were used. 𝜆. 3. z. 𝜑 DoF respectively. y. 10 node. Each simulation result is depicted as a deformed structure to show the mode shape.644kHz for the x. stress-strain module with the Lagrange-quadratic eigenfrequency solver was used to numerically determine several eigenfrequencies. θ. 𝑥: 𝑚𝜆2 + 𝑘𝑖 = 0 𝑦: 𝑚𝜆2 + 𝑘𝑗 = 0 𝑧: 𝑚𝜆2 + 𝑘𝑘 = 0 𝜃: 𝐼𝑗 𝜆2 + 𝑑𝑖2 𝑘𝑘 = 0 𝜑: 𝐼𝑖 𝜆2 + 𝑑𝑗2 𝑘𝑘 ′ = 0 Therefore the natural frequencies of the system are 4. It was assumed that viscous and frictional damping is negligible and that small vibrations and displacements take place.904kHz. which can be found in the table 1 in the appendix. displacement so that the results can readily be compared to experimental fast Fourier transform (FFT) data.523kHz and 8. All three gyroscopes were designed in MEMS Pro (L-Edit) and then imported into COMSOL. 5. 44 .2 Finite Element Analysis The COMSOL solid.

A mesh of 62747 elements with 391287 degrees-of-freedom (DoF) was used to define the structure of the device within COMSOL. The first five simulated mode shapes are shown in the figures below.3. which is suspended above the substrate.2. 45 . necessary in the MetalMUMPS process to facilitate device release from the substrate.1. The structure was imported into COMSOL and geometrically simplified to reduce computation time. The mechanical structure is made from electrodeposited nickel. Nickel electrical pads are also used to anchor the suspended structure to the substrate.3 Thermally-Actuated Gyroscope Simulation Results The thermally-actuated gyroscope design was manufactured using the MetalMUMPS process outlined in section 2. which caused an acceptable bias error of -1.64% to the mass of the device. The holes in the structure. were removed as well as the sensing parallel plate sensors.

Figure 3-8 .669kHz.Thermally-actuated gyroscope simulated 1st mode at 3. 46 .

47 .Thermally-actuated gyroscope simulated 2nd mode at 4.Figure 3-9 .948kHz.

459kHz. 48 .Figure 3-10 .Thermally-actuated gyroscope simulated 3rd mode at 5.

Figure 3-11 .Thermally-actuated gyroscope simulated 4th mode at 6. 49 .923kHz.

Thermally-actuated gyroscope simulated 5th mode at 8.500kHz. The second mode shape causes the proof mass to move in the x- direction. Figure 3-12 . also defined as the driving mode.948kHz and is defined as the sensing mode of the device.293kHz where the proof mass vibrates about the y-axis and likewise. The scale of the z-displacement in each simulation is arbitrary but useful in comparing the relative magnitudes associated with each eigenfrequency. which occurs at 3.459kHz where the proof mass moves along the y-direction. There is a torsional mode at 6. The first mode shape. However. shows the centre proof mass vibrating straight up-and-down in the z-direction. where there is also a slight torsion mode about the x-axis. the frequency of the mode shape is important in predicting the resonant frequencies of the device prototype. This response occurs at 4.669kHz. The simulated results above show five distinct mode shapes of the thermally-actuated gyroscope. at 50 . The third mode occurs at 5.

923 -5. The structure itself is more complex in comparison to the thermal gyroscope since it contains an additional frame. This information is summarized in the table below: Table 3-1 .7% The difference between the simulated and analytical values can be explained by the initial rigidity and fixed end assumptions made for the analytical model.766 3. fourth and fifth modes are the z.066 5.644 8.9% 3rd y 5.669 2.Thermally driven microgyroscopic sensor analytical and simulated summary of results.459 -7. 3.4 Electrostatically-Actuated MetalMUMPS Gyroscope Simulation Results The electrostatically-actuated MetalMUMPS gyroscope was modelled similarly to the thermally- actuated gyroscope. θ and 𝜑 modes in the analytical model.904 4. The CAD model was not simplified in this case. second. y.523 6.6% 2nd x 4.8. therefore a finer mesh of 273378 elements with 1547550 DoF was required to define the increased complexity of the gyroscopes geometry while sacrificing computation time.948 -0. Several of the significant mode shapes are shown below. DoF Analytical Simulated % Error Mode Shape Eigenfrequency (kHz) Result (kHz) 1st z 3. 51 . x. which increases the degrees-of-freedom. third.500kHz the proof mass vibrates about the x-axis.2% 4th θ 6. These mode shapes correspond to the analytically determined mode shapes where the first.8% 5th 𝜑 8.500 1.

Electrostatically-actuated MetalMUMPS gyroscope simulated 1st mode at 1.170kHz. 52 .Figure 3-13 .

Electrostatically-actuated MetalMUMPS gyroscope simulated 2 nd mode at 1.Figure 3-14 . 53 .644kHz.

Electrostatically-actuated MetalMUMPS gyroscope simulated 3rd mode at 1. 54 .Figure 3-15 .941kHz.

Electrostatically-actuated MetalMUMPS gyroscope simulated 4th mode at 2.316kHz. 55 .Figure 3-16 .

Figure 3-17 .246kHz.Electrostatically-actuated MetalMUMPS gyroscope simulated 7th mode at 4. 56 .

which occur at 1. The out-of-plane mode shapes that are shown are the third. second and fourth mode shapes are in-plane modes. which exclusively affects the proof mass since it is the only mass that has a DOF in that direction.510kHz. x-direction. The first.644kHz and 2. Both the outside frame and proof 57 . Figure 3-18 . The third mode shows the proof mass and middle frame moving in-phase with each other while the outside frame remains stationary.941kHz. 4. The fourth mode also vibrates in the driving direction but the middle and outside frames vibrate out of phase with each other. These were selected because they were observed in the experimental data. The first mode is in the driving. seventh and eleventh modes.170kHz. y-direction. which only affects the middle frame.Electrostatically-actuated MetalMUMPS gyroscope simulated 11th mode at 8.510kHz. which occur at 1. The second mode is in the sensing.246kHz and 8. 1.316kHz.

The first seven mode shapes that were derived from the simulated model are shown in the figures below.mass move out-of-phase with each other in the seventh mode shape. 58 . The eleventh mode shape involves the proof mass and outside frame moving out-of-phase with the middle frame. 3. The middle frame bends in response to the outside frame’s displacement but is motionless otherwise. A mesh of 101061 elements with 585126 DoF was used in the simulated model.5 Electrostatically-Actuated SOI Gyroscope Simulation Results The structure of the electrostatically-actuated SOI gyroscope is made of single-crystal silicon as opposed to nickel like the MetalMUMPS devices. The device itself is relatively simple in comparison with only an outside frame and a proof mass with supporting flexures to suspend the structure. The thin-film silicon layer is half the thickness of the MetalMUMPS layer as well at 10µm. which affects the spring and mass constants of the mechanical components of the system.

Electrostatically-actuated SOI gyroscope simulated 1st mode at 2. 59 .159kHz.Figure 3-19 .

Electrostatically-actuated SOI gyroscope simulated 2nd mode at 3. 60 .Figure 3-20 .298kHz.

454kHz.Electrostatically-actuated SOI gyroscope simulated 3rd mode at 3. 61 .Figure 3-21 .

Figure 3-22 . 62 .679kHz.Electrostatically-actuated SOI gyroscope simulated 4th mode at 3.

494kHz. 63 .Figure 3-23 .Electrostatically-actuated SOI gyroscope simulated 5th mode at 5.

Figure 3-24 .500kHz. 64 .Electrostatically-actuated SOI gyroscope simulated 6th mode at 5.

The outside frame vibrates up-and- down in the z-direction at the mode shape that occurs at 5. At 3. At 2. The fifth mode shape figure shows a torional mode where the outside frame vibrates about the y-axis at 5. Figure 3-25 .159kHz the driving mode operates exclusively on the proof mass in the z-direction.476kHz. 65 .476kHz.679kHz there is a torsional mode where the proof mass is rotating about a nodal axis located 45° and 135° in the x-y plane respectively.298kHz. The sensing mode shape acts in the y-direction on the outside frame and occurs at 3.494kHz and the proof mass bending in response to the applied forces from the connecting flexures. The seventh mode shape figure shows the outside frame vibrating about the x-axis at 7.500kHz.Electrostatically-actuated SOI gyroscope simulated 7th mode at 7.454kHz and 3. The first and second modes are defined as the driving and sensing modes respectfully.

The experimental setup for the dynamic testing is demonstrated in Figrue 4. a scanning laser measured the out-of-plane motions of the gyroscope and relayed the information to the Polytec OFV-5000 vibrometer controller. Queen’s University. Using laser interferometry. The original and amplified signals were displayed using a Tektronix TDS 3052 Digital Oscilloscope and the electrical current was measured using a Mastercraft Digital Multimeter. The signal was amplified using an FK602 2-Watt power amplifier.1 Thermally-Actuated Gyroscope Experimental Testing 4. The onboard signal generator was used to supply the actuation signal. The resultant data was converted to microns and plotted. The displacement of the actuator was measured in pixels by the image processing software by comparing each image to a reference image where no current applied. 66 . 212.1.1. Chapter 4 Experimental Determination of Modal Frequencies The purpose of experimentally determining the modal frequencies of the gyroscopes is to verify the simulated models and to predict their operational performance. The original actuation signal is compared to the vibrometer measurements. A diagram showing the relationships between the apparatuses is shown in the diagram in Figure 4. which was powered by an Agilent DC source. Dynamic measurements were made with the Polytech Microsystem Analyzer MSA-400. Individual images of the actuator were taken at different currents supplied by an Agilent DC source. Static measurements were performed using the National Instruments Vision Assistant AutoMax software.2.1 Thermally-Actuated Gyroscope Testing Methodology All testing was conducted in the MEMS Lab in Jackson Hall Rm. 4.

Diagram of the electrical relationships between testing equipment apparatuses used for the thermally-actuated gyroscope. The 67 . Polytec OFV-5000 Polytec OFV-552 Vibrometer Fiber Vibrometer Controller Polytec MSA-400 Tektronix TDS 3052 Micro-gyroscope Junction Box Digital Oscilloscope Agilent DC Power FK602 .2 Watt Mastercraft Supply Power Amplifier Multimeter Figure 4-1 . The Polytec MSA-400 can dynamically characterize in-plane modes using the Planar Motion Analyser (PMA-400) and out-of-plane modes using the Scanning Vibrometer (PSV-400).Polytec MSA-400 and testing setup. Figure 4-2 .

Figure 4-3 . Each image is compared to the original reference image and the distance between the components of interest is measured in pixels between the two images.Polytec PMA reference image example. The PSV uses laser interferometry to determine the velocity of the out-of-plane component of MEMS devices. This is a similar process to the one used for static testing. The displacement of the component being measured is determined by image processing. An example of a reference image is shown in figure 4-3 below.PMA uses stroboscopic images to measure the displacement of MEMS devices. A measurement laser is scattered on the surface of a vibrating device that is then reflected back to the scanning head where the measurement laser interferes with a second 68 . An image is captured at different phase angles within an excitation signal waveform and spliced together to form a video. The green box indicates the searching area and the red box is the area-of-interest.

Laser interferometry equipment diagram20. The phase shift is measured using a photosensor and digitized so an FFT can be produced from the recorded data. Figure 3-1 in the previous chapter shows the boundary conditions of the device. All measurements were taken at atmospheric pressure and room temperature. 69 . Figure 4-4 . Figure 4-5 below shows a scanning electron microscope (SEM) image of the gyroscope with the significant features identified. 4.1. which is outlined in section 2.1.2 Thermally-Actuated Gyroscope Prototype The electro-thermally driven micro-gyroscope was fabricated using the MetalMUMPS process. A visual diagram of this process can be seen in Figure 4-4.2. The testing equipment was isolated from environmental vibration noise using a pneumatically damped table.reference laser.

Decoupling Frame

Comb Sensors

Triple-Folded Proof Mass
Flexures

Double-Folded
Flexures

Chevron Actuator
y

x

Figure 4-5 - SEM image of an electro-thermally actuated MEMS gyroscope prototype.

When the decoupling frame is driven using the chevron actuator and an angular velocity is

applied about the out-of-plane, z-direction then a Coriolis acceleration is applied to the proof

mass in the x-direction, which is then detected by the electrostatic comb sensors.

4.1.3 Thermally-Actuated Gyroscope Static Measurements

The response of the thermal actuator under a static excitation signal was first investigated to

predict the actuation current required for dynamic testing and to provide additional insight on the

behaviour of the actuator. The displacement was determined using the National Instruments

Vision Assistant package. The current was steadily controlled and increased using an Agilent DC

power source. The results of the test can be seen in the graph, figure 4-7, below.

70

6.00

5.00

Displacement (µm)
4.00

3.00

2.00

1.00

0.00
0 45 90 135 180 225 270 315
Current (mA)

Figure 4-6 - Thermal actuator displacement and current relationship.

The error bars indicate the inaccuracy associated with the resolution of the images, which is

limited by the magnification of the lens and number of pixels in each image. A 1 pixel change

corresponds to a 0.3µm displacement, so planar displacements below 0.3μm cannot be

determined accurately using this method.

The thermal actuator exhibits relatively little displacement under 150mA of current. This is

attributed to surface stiction preventing the movement of the gyroscope frame under low thermal

expansion forces. Above 150mA the surface stiction is overcome, which leads to a sharp increase

in displacement followed by a linear relationship above 180mA. Above 300mA the chevron

actuator begins to bow out-of-plane making it difficult to measure the in-plane displacement and

becomes non-linear once again. Current between 180mA-300mA is the target operating range for

the excitation signal of this device.

71

4.1.4 Thermally-Actuated Gyroscope Planar Driving Mode Resonant Frequency

Characterization

A sinusoidal AC 225mARMS excitation signal was applied through the chevron actuator to elicit a

resonant response from the structure. The driving mode natural frequency was identified by

scanning multiple frequencies and analyzing the displacement and quality of the mechanical

vibration. No DC offset was applied to the signal; consequently the frequency of the mechanical

response is twice the frequency of the driving signal due to the squaring of the driving current

waveform from joule heating. The maximum displacement occurred at a forcing frequency of

approximately 3.0kHz with a mechanical response above the noise level between 2.5kHz and

3.5kHz. A comparison of the three frequency responses is shown in figure 4-8 below.

2.00
2.5kHz 3kHz 3.5kHz
1.50

1.00
Displacement (µm)

0.50

0.00
0 45 90 135 180 225 270 315 360
-0.50

-1.00

-1.50

-2.00
Signal Phase (°)

Figure 4-7 - Frequency response using a 225mARMS sinusoidal excitation signal at three frequencies.

The driving resonant frequency, therefore, occurs at a frequency of 6.0kHz where a peak-to-peak

displacement of approximately 3µm is elicited from the structure using a 225mARMS AC signal.

At 5kHz and 7kHz the displacement is significantly less at 0.3µm, however the vibration is still

above the noise level and can still be used for exciting the sensing mass if necessary. The driving

72

50 3. Once the resonant frequency of the driving mode was identified to be 6. the displacement amplitude for changes in current at a constant frequency was investigated. Seventeen separate tests were performed for currents between 0mA to 577mA. A table of the results can be found in table 2 in the appendix. which is graphed in the figure below. The structure was vibrated at an operating frequency of 6. The maximum peak-to-peak displacement was recorded with the RMS AC current and resultant power.50 1.50 4.0kHz.00 1.50 0.50 2.mode resonant frequency is the most efficient operating frequency and would cause the greatest response in the sensing mass if the mode shape frequencies coincided. however greater velocities will induce changes in the damping coefficients and higher temperatures will affect the spring coefficients due to changes in Young’s modulus. otherwise.00 0. which is close to the predicted elastic limits of the actuator. It is assumed that the operating voltage does not influence the resonant frequency of the device.00 2. Displacements amplitudes for RMS currents below 20mA were impossible to extract from the data due to noise. relatively consistent and 73 .00 0 100 200 300 400 500 600 Current (mA) Power (mW) Figure 4-8 .Power consumption of the thermal actuator at several displacements elicited by changes in operational current.0kHz for all tests to induce the greatest response above the noise level. The response from the testing was.00 Displacement (µm) 3. 4.

50 1. This is caused by erroneous measurements made by the image processor when the thermal actuator moves out-of- plane and away from the focal point of the microscope.50 Phase (°) Figure 4-9 . The 298mA signal is subject to more noise in comparison to the 225mA signal. which would be caused by a temperature offset due to insufficient cooling. a third-order polynomial fits the data much better in comparison with an R2 value of 0.997. which can be observed visually as well. However.50 52mA 225mA 298mA 2.00 -1. The 298mA signal is also offset by a displacement 0.50 0. suggesting that there are non-linearity’s present.2. The out-of-plane motions are undesirable due to changes in the apparent area of the capacitive fingers in the sensing circuit. The displacement can be represented with a second-order polynomial.00 Displacement (µm) 0. This is likely caused by energy being leeched by out-of- plane motions of the system such as bowing of the thermal actuator.3. which has an R2 value of 0. The forcing frequency response curves. shown below.25µm.00 -2.00 -0.predictable. Another point-of-interest is the inflection point that occurs at 4.2µm of displacement. provide additional insight on the decrease in displacement from greater power consumption. as predicted by the governing equations shown in section 1. 2.50 0 45 90 135 180 225 270 315 360 -1.00 1. The out-of-plane 74 .Displacement of the thermal actuator at 3kHz excitation for different currents.50 -2.968.

1. The electrostatic parallel plate sensors were used as comb-drive actuators to excite the proof mass in the sensing direction. This was done by applying a 40V AC potential to the sensing fingers and gyroscope frame.displacement currents would not be selected regardless of these detrimental effects because the displacement for the 225mA signal dominates the greater current levels.5 Thermally-Actuated Gyroscope Planar Sensing Mode Resonant Frequency Characterization The sensing mode resonant frequency was identified with the PMA. 4. 75 . A graph of the results can be found figure 4-11 below. A finer sweep was used once the natural frequency was identified to define the resonant peak. To find the resonant frequency a 2kHz to 8kHz frequency sweep range was initially used.

00 -300.00 60.00 -200. 140. At this frequency the amplitude of the displacement is 0.00 200.00 100.00 0.00 20.00 Frequency (Hz) Figure 4-10 – Frequency response of the thermal gyroscope proof mass in the sensing direction.12µm using a 40V AC signal.98kHz for the in-plane sensing mode direction. The relative proximity of the driving and sensing natural frequencies is beneficial to the output signal quality of the device. it may be detrimental to the device’s accuracy due to susceptibility to temperature or pressure changes in the environment. The resonant frequency occurs at 5.00 Phase (°) 0.00 2000 3000 4000 5000 6000 7000 8000 Frequency (Hz) 400. These temperature and 76 .00 100.00 80.00 -400. however.00 40.00 2000 3000 4000 5000 6000 7000 8000 -100.00 300.00 Magnitude Displacement (nm) 120.

The FFT shown is the average spectrum of a matrix of points selected on the proof mass.00 3rd Mode 4th Mode 35.00 25. Images of the mode shapes can be seen in figure 4-13. The system response was then recorded as an FFT to experimentally determine the eigenfrequencies.00 5. 77 .00 20.pressure changes could shift the resonant frequencies of the modes.7V AC signal was applied to the chevron actuator to induce small vibrations in the structure.Experimental out-of-plane FFT of the thermally-actuated gyroscope.6 Thermally-Actuated Gyroscope Out-of-Plane Characterization The out-of-plane response of the proof mass was found using the laser vibrometer.00 1st Mode 15. which would alter the force response due to the Coriolis acceleration acting on the sensing mass.00 40. Figure 4-12 contains the FFT of the out-of-plane velocities caused by small vibrations. A 0. 4. The displacement vector of each point can be derived from the measurements and plotted such that the mode shapes can be reconstructed in a 2D plane.00 0 2000 4000 6000 8000 10000 Frequency (Hz) Figure 4-11 .1. 45.00 0.00 Magnitude (μm/s) 30.00 2nd Mode 5th Mode 10.

The third and fourth peaks.1kHz and 6.2.5kHz and have the largest amplitude.9kHz. both have the same mode shape that rotates about the sensing axis. 78 . which occur at 6.9kHz is a straight up-and-down mode. B) 2nd observed mode at 5.2 Electrostatically-Actuated Gyroscopes Experimental Testing 4.1kHz and 6.Experimentally determined mode shapes for a thermally-actuated gyroscope at A) 1st observed mode at 2.1. at 5. rotates about the driving axis and slightly wobbles about the sensing axis.4kHz. 4.1. C) 3rd and 4th observed mode at 6. A) B) C) D) Figure 4-12 .5kHz. The second mode.1 Electrostatically-Actuated Gyroscopes Testing Methodology Testing for the electrostatically-actuated MetalMUMPS gyroscope was performed similarly to the thermally actuated gyroscope testing as outlined in section 4. The final observed mode shape also rotates about the sensing direction. D) 5th observed mode at 8.6kHz The first mode shape that occurs at 2.4kHz. however voltage amplification was required in place of current amplification. instead the actuation side of the proof mass is the nodal point.

79 .2. The Tabor Electronics Voltage Amplifier had a gain of 50 for the operational frequency range it was used at. Computer Input Polytec MSA-400 Tektronix TDS 3052 Polytec OFV-5000 Junction Box Digital Oscilloscope Vibrometer Controller Tabor Electronics Micro-gyroscope Polytec OFV-552 Wideband Amplifier Fiber Vibrometer 9400 Figure 4-13 .Diagram of the electrical relationships between testing equipment apparatuses used for the electrostatically-actuated MetalMUMPS gyroscope.1. which is outlined in section 2.2.2 Electrostatically-Actuated MetalMUMPS Micro-gyroscope Prototype The electrostatically driven micro-gyroscope was fabricated using the MetalMUMPS process. Figure 4-15 below shows an SEM image of the gyroscope with the significant features identified. 4.

Double-Folded Flexures Triple-Folded Parallel Plate Decoupling Frame Flexures Sensors Proof Mass y Comb Drive Actuator x Figure 4-14 – SEM image of an electrostatically-actuated MetalMUMPS gyroscope prototype. The decoupling frame isolates the Coriolis induced force acting on the proof mass in the x-direction.2. also defined as the sensing direction in this case. 4.3 Electrostatically-Actuated MetalMUMPS Gyroscope Planar Driving Mode Resonant Frequency Characterization The in-plane response of the MetalMUMPS electrostatically-actuated gyroscope was determined using the PMA. A frequency sweep from 1kHz to 8kHz was used to identify the location of the resonant frequencies and a second finer sweep was performed to characterize the frequency 80 . A 25VAC with 25VDC offset signal was applied to the comb drive actuator to excite the proof mass. The gyroscope is driven by comb drive actuators in the y-direction that are located at both ends of the device. The device itself is approximately the same size as the thermally-actuated gyroscope with a footprint of 6mm2 and a thickness of 20µm. The symmetrically placed parallel plate capacitive sensors detect movement in the sensing direction.

both frequency response curves are shown in Figure 4-16 and 4-17. 81 .response curve of both the driving and sensing frequency.Frequency response of the electrostatic gyroscope proof mass in the driving direction. 1200 Magnitude Displacement (nm) 1000 800 600 400 200 0 1300 1400 1500 1600 1700 1800 Frequency (Hz) 210 180 150 120 Phase (°) 90 60 30 0 1300 1400 1500 1600 1700 1800 -30 Frequency (Hz) Figure 4-15 .

A third in-plane mode was identified by the simulated results.56kHz and 1.91kHz respectively from the experimental data shown above. and was detected during the initial frequency sweep as well. labelled as the fourth mode in section 3.Frequency response of the electrostatic gyroscope proof mass in the sensing direction. 1600 Magnitude Displacement (nm) 1400 1200 1000 800 600 400 200 0 1700 1800 1900 2000 2100 2200 2300 Frequency (Hz) 0 1700 1800 1900 2000 2100 2200 2300 -50 -100 Phase (°) -150 -200 -250 -300 -350 Frequency (Hz) Figure 4-16 . The identification of the 82 . However. the out-of- phase response of the driving frame and the proof mass made it difficult to identify the location of the resonant frequency due to interference of the two mass’s vibration.3. The driving and sensing mode resonant frequencies were determined to be 1.

Between 0kHz and 20kHz there are five significant peaks that were identified.00 4th Mode 5. 4.00 3rd Mode Magnitude (μm/s) 20.00 0 5000 10000 15000 20000 Frequency (Hz) Figure 4-17 . The entire non-fixed gyroscope structure was scanned to identify all the significant mode shapes on all three frames. The experimentally derived mode shapes can be seen in figure 4-19. 83 . 30.4 Electrostatically-Actuated MetalMUMPS Gyroscope Out-of-Plane Characterization The gyroscope was driven using a 30VAC 1kHz .resonant frequency of the fourth mode is not necessary to determine the ideal operational frequency of the device and can be ignored.Experimental out-of-plane FFT of the electrostatically-actuated gyroscope. The out-of-plane mechanical response of the system was measured using the PSV system to determine the experimental eigenfrequencies.2.00 10. The resultant FFT is shown in figure 4-18.00 5th Mode 0.40kHz burst chirp excitation signal with 30VDC offset.00 2nd Mode 1st Mode 25.00 15.

1kHz and D) an out-of-plane mode at 12. A) B) C) D) Figure 4-18 . B) 7th mode at 7.3kHz.Experimentally determined mode shapes for a thermally-actuated gyroscope at A) 3rd mode at 2.9kHz 84 .7kHz. C) 11th mode at 10.

The in-plane dimensions of the SOI and MetalMUMPS devices are comparable.3kHz. 85 . 4.9kHz is a wobbling vibration that only acts on the outer frame. The peak that occurs at 10. The mode shape that occurs at 12. The second mode shape is similar except that the proof mass vibrates 180° out-of-phase with the decoupling and outside frame at 7.2.7kHz. The device has an approximate thickness of 10µm. however the SCSi layer of the SOI device does not show up clearly in SEM imaging.2.5kHz are not shown due to noise interfering with the interpretation of the mode shape.The first mode shape.3 Electrostically-Actuated SOI Gyroscope Prototype The electrostatically driven micro-gyroscope was fabricated using the SOI process. which occurs at 2. however the proof mass and outside frame are 180° out-of-phase with the decoupling frame.1kHz is also an up-and-down mode shape. is a straight up-and-down mode that acts on the proof mass and decoupling frame. which is outlined in section 2. The images of the peak that occurs at 15. The figure below shows an SEM image of the SOI design made with the MetalMUMPS process.

Testing was not possible due to the lack of functioning prototypes. Capacitive plate sensors underneath the outside frame detect the displacement in the y-direction. A rotation about the x-axis will induce an acceleration in the y-direction when the proof mass is actuated. 86 . The thin cantilever beams were broken. The 1μm thin cantilever beams ideally isolate the out-of-plane motion from the surrounding frame. Double-Folded Flexures Proof Mass Cantilever Beams y x Figure 4-19 – SEM image of an electrostatically-actuated MetalMUMPS gyroscope prototype: SOI design The operational principles of the out-of-plane device are different compared to the other two designs. which was likely due to fabrication complications or the structure may have been too weak to survive transport. The proof mass is actuated out-of-plane in the z-direction by a capacitive plate concealed underneath the mass.

The driving frame and decoupling frame resonate out-of-phase with each other. 5. Other mode shapes can potentially interfere with the sensing performance of the gyroscopes. 87 . however the fourth simulated mode. The same problem affects the electrostatically-actuated MetalMUMPS gyroscope. found in figure 3-10 in section 3. However this must be accounted for in the original design by separating the sensing mode resonant frequency from any unfavourable mode shape resonant frequencies. Once identified. One method that can be used to mitigate the error associated with detrimental mode shapes. the amplitudes of the mode shapes can be reduced by avoiding the resonant frequencies of the detrimental modes when exciting the device. The detrimental mode shapes can be predicted using simulated modal analysis and experimental evaluation.1 Detrimental Mode Shapes The only two desirable mode shapes are the driving and sensing modes for each gyroscope. which reduces the driving velocity on the sensing frame that will decrease the strength of the output signal. The out-of-plane data for the thermally-driven gyroscope in section 4. The proof mass vibrates out-of-plane causing error due to the reduction in the apparent area of the parallel sensing plates from any displacement in the z-direction. Chapter 5 Simulated and Experimental Results Analysis This section explores the significance of the results reported in chapters 3 and 4.6 shows several modes that are detrimental to the sensing performance of the device. interferes in-plane.3. which is why each gyroscope uses symmetrical capacitive sensors to reduce error from sub-optimal mode shape displacements.

1 Thermally-Actuated Micro-gyroscope The experimental data matched the order of the simulated modes but there is a large discrepancy in the resonant frequencies where they occur. 88 . 5. This information is summarized in the table below. but this would be impossible with the current testing setup. Ideally. The PSV would then be used to produce an FFT of the proof mass and the magnitude of the displacement would be determined.The error associated with out-of-plane mode shapes can be determined while driving the gyroscope at its operational frequency.2. an angular velocity would be applied as well.2 Comparison of Simulated and Experimental Results The results from the simulations and experimental tests are compared to determine the accuracy of the simulated models and to give additional insight to the interpretation of the experimental data. 5.

0 4. 6.Thermally-actuated gyroscope mode shape comparison.293 6.9% 39. 8.2. 89 .500 6.4 -8.2 Electrostatically-Actuated MetalMUMPS Micro-gyroscope The out-of-plane mode shapes for the electrostatically-driven MetalMUMPS gyroscope also occur in order according to the simulated model.5kHz can be explained by the in-plane driving mode exciting the out-of-plane mode such that there are two observable peaks with the same mode shape.0 -9. The data for the out-of-plane modes for both the simulated and experimental data and the in-plane modes for the simulated data are shown below. 5.1kHz and 6.948 5.3%. Table 5-1 .669 2.4% 5.0% 6. Mode Shape Simulated (kHz) Experimental (kHz) Error (%) 3.8% The experimental double-peak at 6.1.9 26.5 30.5% 4.459 6.

941 2.1 -15.6 -26.644 1.510 10.7% 90 .316 Not Available Available 4.9 -13.5% 1.3 -15. Mode Shape Simulated (kHz) Experimental (kHz) Error (%) 1.9% 1.6% Not 2.170 1.246 7.Table 5-2 .Electrostatically-actuated MetalMUMPS gyroscope mode shape comparison.9% 8.7 -44.

The quality factors and damping coefficients for each experimentally determined resonant peak are shown in figures 1 and 2 in the appendix. Higher-energy modes typically overwhelm those with smaller amplitudes as well. the damping coefficient. upon calculating the quality factors from the experimental data using equation 1. The information contained in table 5-1 is agreeable with this theory. The material properties are also expected to be inconsistent within the gyroscope structure depending on imperfections and the size of the grain boundaries.8 and determining the damping coefficients using equation 1.2. therefore damping can be accurately neglected from the simulated model. for each mode shape was determined to be much less than 0. The damping factor is also expected to vary with each mode shape depending on the contribution of viscous shearing forces and pressure differentials across the vibrating element.Due to the increased relative complexity of the electrostatically-actuated MetalMUMPS gyroscope. The material properties from the COMSOL library and those of the gyroscopes can vary and cause a disparity between the simulated and experimental data.7. This is expected to reduce the resonant frequencies in each mode compared to the simulated eigenfrequencies. An in-plane mode will be less damped compared to an out-of-plane mode because there will be a greater shear force and a smaller pressure differential due to the surface area in each plane.3 Error Explanation There are several factors that can explain the discrepancies found between the simulated and experimental results. ζ.1. it has a high mode shape density making it difficult to differentiate between modes. It is expected that the experimental in-plane resonant frequencies will be closer to the simulated values for this reason. which makes them difficult to identify. 5. However. The two significant properties that would affect the difference between the experimental and simulated models are density and Young’s modulus. One source of error is the exclusion of damping from the simulated model. The values of 91 .

The thickness of MetalMUMPS structures can vary up to 20±10µm. 92 . the simulated and prototype dimensions are identical except for alterations made to the thermal gyroscope to simplify the structure and dimension variation due to fabrication tolerances. which can result in large discrepancies between the actual device and the theoretical mass and spring matrices. These variations are expected to change the spring and mass values for each element within the gyroscope’s structure. Therefore any difference in the actual value of the Young’s modulus would not account for all the error reported in tables 5-1 and 5-2. According to the analytical model.the properties are dependent on the type of fabrication method used and the thickness of the thin film. The dimensions of the gyroscopes are also expected to vary between the simulated models and the prototypes. The CAD models. which were used for fabrication. an inherent limitation of the MetalMUMPS process. would result in a discrepancy of 5.11 A thickness of 30µm would result in a discrepancy between the simulated and experimental results up to 50% according to the analytical model. which would account for all the error in the out-of-plane results.8%. Any deviation of the actual material properties from the default COMSOL properties is expected to shift the resonant frequencies that the mode shapes occur at. The spring values are more susceptible to changes in the dimensions because of the relative thinness of the flexures. The values that were used in the simulations are contained in table 1 in the appendix. were imported directly into COMSOL. a difference of 12% between the simulated and experimental value. Two other common causes of dimensional error include slanted side-walls. The range of values for the Young’s modulus of nickel has been reported by various sources to be anywhere between 150GPa and 245GPa21. and mushrooming of the electroplated nickel. a Young’s modulus of 245GPa. which were taken directly from the COMSOL material library.

They are not particularly significant in bottom-up fabrication processes such as electroplating but can still affect the resonant frequencies of a mechanical system.Residual stresses may also be present within the structure after fabrication. 93 .

Three in-plane modes were identified in the simulated model. The gyroscope is expected to have poor temperature and pressure sensitivity because of the high quality factor of the sensing and driving mode frequency response curves. the out-of-phase driving mode. Sensitivity is one of the performance indicators of gyroscopic sensors. Having both modes directly coincide is ideal if the objective is to elicit the greatest response using the least amount of energy. The gyroscope also will not take advantage of mode coupling because of the relatively large difference in proximity between the driving and sensing modes.3. and at 1.1 Thermally-Actuated MetalMUMPS Gyroscope Mode Matching The in-plane analysis in chapter 4 for the thermally-actuated gyroscope indicates that the driving and sensing modes are closely matched.3 Performance Analysis 5. estimated to occur at 2. The velocity of the proof mass is reduced by 94 . however the system’s performance is relatively unstable in this case. The driving mode occurred at 6.316kHz.2 Electrostatically-Actuated MetalMUMPS Gyroscope Predicted Mode Matching Experimental in-plane mode testing has not been performed. which will need to be compensated by additional circuitry.5. will negatively affect the performance. Fluctuations in temperature and pressure can affect the sensing mode’s resonant frequency and change the magnitude of the response that will ultimately affect the output.98kHz.9kHz the proof mass resonates in the sensing direction.3. At 1. 5. It is expected that the thermally-actuated gyroscope will be highly dependent on environmental temperatures and pressures during operation. Also. consequently the simulated data is used to predict the modal response of the electrostatically-actuated gyroscopes.6kHz the proof mass and decoupling frame vibrate in the driving direction. The sensitivity of environmental conditions can be tested by applying an angular velocity and changing the temperature and pressure then measuring the output.0kHz and the sensing mode occurred at 5.

It was not possible to perform tests using a square wave actuation signal due to limitations of the testing equipment.4 Thermally-Actuated Gyroscope Power and Efficiency Electrical power tests and the amplitude response from the gyroscope. 5.298kHz respectively. are important to the viability of thermally-actuated gyroscope designs.the out-of-phase motion. A significant disadvantage of electrostatically-actuated designs is the high voltage requirement. which requires on-chip circuitry and can sometimes restrict design possibilities. however several adjustments can be made to drive the gyroscope with a microprocessor.3. therefore excitation signals close to this frequency should be avoided during operation.3. Thermal actuators require much lower voltages that make them easily implemented with digital CMOS circuitry. 5. Microprocessors cannot output this much current. summarized in figure 4-8.2µm. however a square wave driving signal would be a better waveform to use because of its greater power 95 . The largest amplitude observed was 4. Increasing the power only induced out-of-plane responses. Microprocessors are only capable of outputting square wave signals as opposed to the sinusoidal actuation signals were used during testing to drive the gyroscope. which is evident in the power tests that were performed. The driving response of the gyroscope will also be relatively small because of the large disparity between the driving and sensing modes. The disadvantage of thermal actuators is their large power requirements compared to electrostatic actuators. which will affect the response in the sensing direction.39mW to achieve. which required 363.3 Electrostatically-Actuated SOI Gyroscope Predicted Mode Matching The simulated driving and sensing mode resonant frequencies was determined to be 2. The frequency span between both modes may be too great for the gyroscope to take advantage of mode matching.159kHz and 3. however it is difficult to speculate without knowing the quality factor of each peak.

a 1. which would also elicit a faster response. 96 . There is energy loss in the wires with the current testing setup that would be reduced if shorter interconnects were used.density. however the gyroscope would have to be designed and fabricated around these specifications.2µm is not necessary. Future thermally-actuated gyroscopes could incorporate this in their design specifications to optimize the efficiency. The power requirements of the gyroscope would be further reduced by using on-chip circuitry. Currently. the power used to elicit a peak-to-peak displacement of 4. Further testing is required to determine how low the driving current can be while still maintaining an acceptable output signal. It would be more efficient if a 5V drop was observed. Adjusting the voltage drop across the thermal actuator is one prerequisite for testing using a microprocessor.5V RMS drop is observed. albeit at a reduced response due to the decrease in driving velocity. A smaller electrical current will also excite the proof mass in the sensing direction as well. Also. most microprocessors output 5V.

Similarly. The natural frequency of the driving and sensing modes of the thermally-actuated gyroscope was experimentally 97 . Chapter 6 Conclusion The objective was to determine and characterize the modal behaviour of three MEMS gyroscopic sensors using simulated models in COMSOL and through experimental methods. Simulated models were created for all three gyroscopes from CAD files to reduce the potential for dimensional error. the electrostatic actuator required a 25VAC with 25VDC offset to actuate. which were then used as predictions for the experimental PSV and PMA testing to help interpret the results. The frequency response of the in-plane sensing mode was also measured by using the parallel plate sensing fingers as actuators by applying a 40V AC driving signal. Several mode shape resonant frequencies were solved for in each model. Each gyroscope had its own distinguishing features. The maximum peak-to-peak displacement observed was 4. one was a novel thermally-actuated design. The thermal actuator required a driving current of at least 150mA to elicit a significant response.2µm at 225. Experimental testing was then performed for the thermally-actuated gyroscope and electrostatic MetalMUMPS gyroscope. Seventeen dynamic tests were performed to characterize the response of the thermal gyroscope at different driving currents. Static testing was performed first to identify the driving signal power requirements. The frequency response of the thermal actuator was determined similarly by measuring the amplitude of the displacement and varying the frequency. another used an electrostatic actuation method. Increasing the current provoked out-of-plane responses and reduced the in-plane displacement. which was manufactured using the MetalMUMPS fabrication process and another was an out-of-plane SOI design.4mARMS current.

the locations of the natural frequencies of the driving and sensing modes were used to predict the response of the gyroscope from mode matching.1 Future Work Recommendations The simulated models need additional work if they are to be useful for the fabrication of future gyroscope models.determined to be 6. Similarly.9kHz respectively. The results were compared to the eigenfrequencies found in the simulated solutions and the discrepancies were attributed to fabrication and material tolerances. model dimensions and residual stresses in the current COMSOL models such that they closely match the experimental data will help with future simulations for other gyroscopes as well if the same settings are used. residual stresses and the exclusion of damping from the simulated model.0kHz and 5.6kHz and 1. Completion of testing on the SOI electrostatically-actuated gyroscope would be useful for revising the simulations. however it will be difficult to derive an accurate viscous damping model that accounts for the variability in damping coefficients for each mode shape. The fabrication and material tolerances were determined to be the most significant contributors to the error between the experimental and theoretical results. An analysis on each significant mode shape helped identify modes that may interfere with the sensing performance of the gyroscopes. Tweaking the material property settings. Similarly. The out-of-plane mode shape resonant frequencies were found using an FFT. the driving and sensing mode resonant frequencies for the MetalMUMPS electrostatically-actuated gyroscope are 1.98kHz respectively. An analytical model as well as quality factors derived from the experimental results was used to determine the most significant contributor to the error that was observed. 6. 98 . The inclusion of damping in a frequency response simulation would be useful for predicting the modal behaviour of the gyroscopes.

The significance of the modal behaviour of each gyroscope will be made explicit by measuring the performance indicators using rate table testing. 99 .2 Thesis Contribution All the simulated and experimental results were original work performed by the author. a previous graduate student at Queen’s University. Insight gained from these tests will be useful in designing future gyroscope devices. 6. The gyroscopic sensors were designed by Rana Iqtidar Shakoor. It is recommended that a new thermal-actuated gyroscope should be designed and fabricated to take advantage of the increased efficiency and ease-of-integration benefits of a thermal actuator that is intended to accept a 5V square wave driving signal.Circuitry for the capacitive sensing plates must be designed and connected as well in order to convert the capacitance to an output voltage for all gyroscopes.

References

1
R. Feynman, “There’s Plenty of Room at the Bottom”, Journal of Micromechanical Systems

Vol. 1, No. 1, March 1992, pp.60-66.
2
M. Dahleh and W. Thomson, “Theory of Vibration with Applications”, 5 th Edition, Prentice-Hall Inc,

1998, ISBN 0-13-651068-X.
3
B. Zhang and D. Fang, “Modeling and modification of the parallel plate variable MEMS

capacitors considering deformation issue”, Mechanism and Machine Theory, 44(4), 2009, pp.

647-655.
4
R. Johnson, “MEMS finds niche in space exploration”, EE Times, 2007

http://www.eetasia.com/ART_8800479803_480500_NT_ca7d6a67.HTM.
5
W. Huang and G. Lu, “Analysis of lateral instability of in-plane comb drive MEMS actuators

based on a two-dimensional model”, Sensors and Actuators A: Physical 113(1), 2004, pp. 78-85.
6
J.-C. Chiou, Y.-J. Lin and C.-F. Kuo, “Extending the traveling range with acascade electrostatic

comb-drive actuator”, J. Micromech. Microeng. 18 (2008) 015018 (7pp).
7
“Lithography”, http://www.memsnet.org/mems/processes/lithography.html.
8
J. Ravnkilde, A. Yalcinkaya, L. Johansen and O. Hansen, “Fabrication of electroplated nickel

micromechanical comb resonators”, Technical University of Denmark DTU, Electrochemical

Society Proceedings Vol. 99-34, pp. 197-202.
9
J. Jahns and K. Brenner, “Microoptics: From technology to applications”, New York : Springer,

2004.
10
O. Geschke, H. Klank, P. Telleman, “Microsystem Engineering of Lab-on-a-Chip Devices”,

Ch. 6, Silicon and Cleanroom Processing (Lecture 2), Wiley-VCH, 2004.
11
“MetalMUMPs Design Handbook”, Revision 2.0, MEMSCAP Inc.
12
“Micragem Data Sheet”, Micralyne Inc.

100

13
B. Bokenhorn and P. Greiff, “A vibratory micromechanical gyroscope”, Navigation and

Control Conference, Minneapolis, MN, Aug. 15-17, 1988, Technical Papers. Part 2 (A88-50160

21-08). Washington, DC, American Institute of Aeronautics and Astronautics, 1988, p. 1033-

1040.
14
Roylance, L. M., and J. B. Angell, “A Batch-Fabricated Silicon Accelerometer,” IEEE Trans.

Electron Devices, ED-26, 1979, pp. 1911–1917.
15
K. Tanaka, Y. Mochida, S. Sugimoto, K. Moriya, T. Hasegawa, K. Atsuchi, and K. Ohwada,

“A micromachined vibrating gyroscope”. Proc., Eighth IEEE Int. Conf. on Micro Electro

Mechanical Systems (MEMS ’95), Amsterdam, Netherlands, Jan. 1995, pp. 278-281.
16
T. Juneau, A. P. Pisano, and J. H. Smith, “Dual axis operation of a micromachined rate

gyroscope” Proc., IEEE 1997 Int. Conf. on Solid State Sensors and Actuators (Tranducers ’97),

Chicago, June 1997, pp. 883-886.
17
Y. Mochida, M. Tamura, and K. Ohwada, “A micromachined vibrating rate gyroscope with

independent beams for the drive and detection modes.” Proc. Twelfth IEEE Int. Conf. on Micro

Electro Mechanical Systems (MEMS ’99), Orlando, FL, Jan. 1999, pp. 618-623.
18
C. Acar, A. M. Shkel “Inherently Robust Micromachined Gyroscopes with 2-DoF Sense Mode

Oscillator” Journal of Microelectromechanical System, Vol 15, No.2, April 2006, pp 380-387.
19
S. Bazaz, M. Burnie, M. Hasan, Y. Lai and R. Shakoor, “Electrothermally Actuated Resonant

Rate Gyroscope Fabricated using the MetalMUMPs”, unpublished.
20
“Polytec Scanning Vibrometer: Theory Manual”, Revision 8.4, Polytec Inc.
21
D.-C. Baek, T.-S. Park and S.-B. Lee, “Measurement of Mechanical Properties of Electroplated

Nickel Thin Film”, Key Engineering Materials Vols. 261-263, 2004, pp 417-422.

101

Appendix

Table Appendix-5-Error! Main Document Only. - Values for material properties used for simulation
models.

Material Young`s Modulus (GPa) Density (kg/m3)
Nickel 219 8900
Single-Crystal Silicon 200 2330

Table Appenix-5-Error! Main Document Only. - Thermally-actuated gyroscope dynamic testing
results.

SG P-P Gyro P-P Gyro RMS RMS Current Gyro Power DC Current Peak-to-
Volt. (V) Volt. (V) Volt. (V) (mA) (mW) (mA) Peak (µm)
0.13 0.13 0.09 0.45 0.04 4
0.32 0.28 0.20 16.68 3.30 11
0.52 0.42 0.30 32.95 9.79 19
0.71 0.57 0.40 48.9 19.71 27 0.17
0.93 0.74 0.52 65.4 34.22 35 0.26
1.13 0.9 0.64 82.1 52.25 43 0.32
1.3 1.08 0.76 98.9 75.53 52 0.38
1.51 1.25 0.88 116.5 102.97 60 0.44
1.72 1.44 1.02 129.4 131.76 67 0.57
1.9 1.65 1.17 138.1 161.12 71 0.63
2.1 1.83 1.29 160 207.04 80 1.00
2.3 1.97 1.39 175.5 244.47 92 1.45
2.51 2.22 1.57 201.6 316.47 104 2.34
2.7 2.28 1.61 225.4 363.39 117 4.20
2.9 2.5 1.77 243.8 430.98 124 4.10
3.1 2.6 1.84 278.3 511.65 138 3.47
3.29 2.74 1.94 297.6 576.59 151 2.70

102

Figure Appendix. Figure Appendix. .Error! Main Document Only.Error! Main Document Only.Experimentally calculated quality factors and damping coefficients of the electrostatically-actuated MetalMUMPS gyroscopic sensor. 103 . – Experimentally calculated quality factors and damping coefficients of the thermally-actuated gyroscopic sensor.